Humans + AI

"Let technology do the bits that technology is really good at. Offload to it. Then over-index and over-amplify the human skills we should have developed over the last 10, 15, or 20 years."

– Kieran Gilmurray

Kieran Gilmurray is CEO of Kieran Gilmurray and Company and Chief AI Innovator of Technology Transformation Group. He works as a keynote speaker, fractional CTO and delivering transformation programs for global businesses. He is author of three books, most recently Agentic AI. He has been named as a top thought leader on generative AI, agentic AI, and many other domains.

Website:

Kieran Gilmurray

X Profile:

Kieran Gilmurray

LinkedIn Profile:

Kieran Gilmurray

BOOK: Free chapters from Agentic AI by Kieran Gilmurray

Chapter 1 The Rise of Self-Driving AI 

Chapter 2: The Third Wave of AI 

Chapter 3 - Agentic AI Mapping the Road to Autonomy

Chapter 4- Effective AI Agents

Understanding the leap from generative to agentic AI

Redefining work with autonomous digital labor

The disappearing need for traditional junior roles

Augmenting human cognition, not replacing it

Building emotionally intelligent, tech-savvy teams

Rethinking leadership in AI-powered organizations

Designing adaptive, intelligent businesses for the future

People

John Hagel

Peter Senge

Ethan Mollick

Technical & Industry Terms

Agentic AI

Generative AI

Artificial intelligence

Digital labor

Robotic process automation (RPA)

Large language models (LLMs)

Autonomous systems

Cognitive offload

Human-in-the-loop

Cognitive augmentation

Digital transformation

Emotional intelligence

Recommendation engine

AI-native

Exponential technology

Intelligent workflows

Ross Dawson: Hey, it's fantastic to have you on the show.

Kieran Gilmurray: Absolutely delighted, Ross. Brilliant to be here. And thank you so much for the invitation, by the way.

Ross: So agentic AI is hot, hot, hot, and it's now sort of these new levels of how it is we — these are autonomous or semi-autonomous aspects of AI. So I want to really dig into — you've got a new book out on agentic AI, and particularly looking at the future of work. And particularly want to look at work, so amplifying cognition.

So I want to start off just by thinking about, first of all, what is different about agentic AI from generative AI, which we've had for the last two or three years, in terms of our ability to think better, to perform our work better, to make better decisions? So what is distinctive about this layer of agentic AI?

Kieran: I was going to say, Ross, comically, nothing if we don't actually use it. Because it's like all the technologies that have come over the last 10–15 years. We've had every technology we have ever needed to make more work, more efficient work, more creative work, more innovative, to get teams working together a lot more effectively.

But let's be honest, technology's dirty little secret is that we as humans very often resist. So I'm hoping that we don't resist this technology like the others we have slowly resisted in the past, but they've all come around to make us work with them.

But this one is subtly different. So when you say, look, agentic AI is another artificial intelligence system. The difference in this one — if you take some of the recent, what I describe as digital workforce or digital labor, go back eight years to look at robotic process automation — which was very much about helping people perform what was meant to be end-to-end tasks.

So in other words, the robots took the bulky work, the horrible work, the repetitive work, the mundane work and so on — all vital stuff to do, but not where you really want to put your teams, not where you really want to spend your time. And usually, all of that mundaneness sucked creativity out of the room.

You ended up doing it most of the day, got bored, and then never did the innovative, interesting stuff.

Agentic is still digital labor sitting on top of large language models. And the difference here is, as described, is that this is meant to be able to act autonomously. In other words, you give it a goal and off it goes with minimal or no human intervention. You can design it as such, or both.

And the systems are meant to be more proactive than reactive. They plan, they adapt, they operate in more dynamic environments. They don't really need human input. You give them a goal, they try and make some of the decisions.

And the interesting bit is, there is — or should be — human in the loop in this. A little bit of intervention.

But the piece here, unlike RPA — that was RPA 1, I should say, not the later versions because it's changed — is its ability to adapt and to reshape itself and to relearn with every interaction.

Or if you take it at the most basic level — you look at a robot under the sea trying to navigate, to build pipelines. In the past, it would get stuck. A human intervention would need to happen. It would fix itself.

Now it's starting to work itself out and determine what to do. If you take that into business, for example, you can now get a group of agentic agents, for example, to go out and do an analysis of your competitors.

You can go out and get it to do deep research — another agentic agent to do deep research, McKinsey, BCG or something else. You can get another agent to bring that information back, distill it, assemble it, get an agent to create it, turn that into an article. Get another agent to proofread it. Get another agent to pop it up onto your social media channels and distribute it.

And get another agent to basically SEO-optimize it, check and reply to any comments that anyone's making. You're sort of going, "Here, but that feels quite human." Well, that's the idea of this.

Now we've got generative AI, which creates. The problem with generative AI is that it didn't do. In other words, after you created something, the next step was, well, what am I going to do with my creation?

Agentic AI is that layer on top where you're now starting to go, "Okay, not only can I create — I can decide, I can do and act." And I can now make up for some of the fragility that exists in existing processes where RPA would have broken.

Now I can sort of go from A to B to D to F to C, and if suddenly G appears, I'll work out what G is. If I can't work it out, I'll come and ask a person. Now I understand G, and I'll keep going forever and a day.

Why is this exciting — or interesting, I should say? Well-used, this can now make up for all the fragility of past automation systems where they always got stuck, and we needed lots of people and lots of teams to build them.

Whereas now we can let them get on with things.

Where it's scary is that now we're talking about potential human-level cognition. So therefore, what are teams going to look like in the future? Will I need as many people? Will I be managing — as a leader — managing agentic agents plus people?

Agentic agents can work 24/7. So am I, as a manager, now going to be expected to do that?

Its impact on what type of skills — in terms of not just leadership, but digital and data and technical and everything else — there's a whole host of questions. There is as much as there is new technology here Ross.

Ross Dawson: Yeah, yeah, absolutely. And so, I mean, those are some of the questions, though, I want to, want to ask you the best possible answers we have today.

And in your book, you do emphasize this is about augmenting humans. It is around how it is we can work with the machines and how they can support us, and human creativity and oversight being at the center.

But the way you've just laid out, there's a lot of what is human work, which is overlap from what you've described.

So just at a first step, thinking about individuals, right? Professionals, knowledge workers — and so they have had, there's a few layers. You've had your tools, your Excels. You've had your assistants which can go and do tasks when you ask them. And now you have agents which can go through sequences and flows of work in knowledge processes.

So what does that mean today for a knowledge worker who is starting to have, where the enterprise starts to bring them in? Or they say, "Well, this is going to support it." So what are the sorts of things which are manifest now for an individual professional in bringing these agentic workforce play? What are the examples? What are ways to see how this is changing work?

Kieran Gilmurray: Yeah, well, let's dig into that a little bit, because there's a couple of layers to this.

If you look at what AI potentially can do through generative AI, all of a sudden, the question becomes: why would I actually hire new trainees, new labor?

On the basis that, if you look at any of the studies that have been produced recently, then there's two roles, two setups. So let me do one, which is: actually, we don't need junior labor, because junior labor takes a long time to learn something.

Whereas now we've got generative AI and other technologies, and I can ask it any question that I want, and it's going to give me a pretty darned good answer.

And therefore, rather than having three and four and five years to train someone to get them to a level of competency, why don't I not just put in agentic labor instead? It can do all that low-ish level work, and I don't need to spend five years learning. I immediately have an answer.

Now that's still under threat because the technology isn't good enough yet. It's like the first scientific calculator version — they didn't quite work. Now we don't even think about it.

So there is a risk that all of a sudden, agentic AI can get me an answer, or generative AI can get me an answer, that previously would have taken six or eight weeks.

Let me give you an example.

So I was talking to a professor from Chicago Business School the other day, and he went to one of his global clients. And normally the global client will ask about a strategy item. He would go away — him and a team of his juniors and equals would research this topic over six or twelve weeks. And then they would come back with a detailed answer, where the juniors would have went round, done all the grunt work, done all the searching and everything else, and the seniors would have distilled it off.

He went — he's actually written a version of a GPT — and he's fed it past strategy documents, and he fed in the client details.

Now he did this in a private GPT, so it was clean and clear, and in two and a half hours, he had an answer.

It literally — his words, not mine — he went back to the client and said, "There you go. What do you think? By the way, I did that with generative AI and agentics."

And they went, "No, you didn’t. That work’s too good. You must have had a team on this."

And he said, "Literally not." And he's being genuine, because I know the guy — he'd put his reputation on it.

So all of a sudden, now all of those roles that might have existed could be impacted.

But where do we get then the next generation of labor to come through in five and six and ten years' time?

So there's going to be a lot of decisions need made. As to: look, we've got Gen AI, we've potentially got agentic AI. We normally bring in juniors over a period of time, they gain knowledge, and as a result of gaining knowledge, they gain expertise. And as a result of gaining expertise, we get better answers, and they get more and more money.

But now all of Gen AI is resulting in knowledge costing nothing.

So where you and I would have went to university — let's say we did a finance degree — that would have lasted us 30 years. Career done. Tick.

Now, actually, Gen AI can pretty much understand, or will understand, everything that we can learn on a finance degree, plus a politics degree, plus an economics degree, plus, plus, plus — all out of the box for $20 a month.

And that's kind of scary.

So when it comes to who we hire, that opens up the question now: do we have Gen AI and agentic labor, and do we actually need as many juniors?

Now, someone's going to have to press the buttons for the next couple of years, and any foresighted firm is going to go, "This is great, but people plus technology actually makes a better answer." I just might not need as many.

So now, when it comes to the actual hiring and decision-making — as to how am I going to construct my labor force inside of an organization — that's quite a tricky question, if and when this technology, Gen AI and agentics, really ramps through the roof.

Ross Dawson: I mean, these are — I mean, I think these are fundamentally strategic choices to be made. As in, you — I mean, it's, crudely, it's automate or augment.

And you could say, well, all right, first of all, just say, "Okay, well, how do we automate as many of the current roles which we have?" Or you can say, "Oh, I want to augment all of the current roles we have, junior through to senior."

And there's a lot more subtleties around those strategic decisions. In reality, some organizations will be somewhere between those two extremes — and a lot in between.

Kieran Gilmurray: 100%. And that's the question. Or potentially, at the moment, it's actually, "Why don't we augment currently?"

Because the technology isn't good enough to replace. And it isn't — it still isn't.

And no, I'm a fan of people, by the way — don't get me wrong. So anyone listening to this should hear that. I believe great people plus great technology equals an even greater result.

The technology, the way it exists at the moment, is actually — and you look at some research out from Harvard, Ethan Mollick, HBR, Microsoft, you name it, it's all coming out at the moment — says, if you give people Gen AI technology, of which agentic AI is one component:

"I'm more creative. More productive. And, oddly enough, I'm actually happier."

It's breaking down silos. It's allowing me to produce more output — between 10 to 40% — but more quality output, and, and, and.

So at the moment, it's an augmentation tool. But we're training, to a degree, our own replacements.

Every time we click a thumbs up, a thumbs down. Every time we redirect the agentics or the Gen AI to teach it to do better things — or the machine learning, or whatever else it is — then technically, we're making it smarter.

And every time we make it smarter, we have to decide, "Oh my goodness, what are we now going to do?" Because previously, we did all of that work.

Now, that for me has never been a problem. Because for all of the technologies over the decades, everybody's panicked that technology is going to replace us.

We've grown the number of jobs. We've changed jobs.

Now, this one — will it be any different?

Actually — and why I say potentially — is you and I never worried, and our audience never worried too much, when an EA was potentially automated. When the taxi driver was augmented and automated out of a job. When the factory worker was augmented out of a job.

Now we've got a decision, particularly when it comes to so-called knowledge work. Because remember, that's the expensive bit inside of a business — the $200,000 salaries, the $1 million salaries.

Now, as an organization, I'm looking at my cost base, going, "Well, I might actually bring in juniors and make them really efficient, because I can get a junior to be as productive as a two-year qualified person within six months, and I don't need to pay them that amount of money."

And/or, actually, "Why don’t I get rid of my seniors over a period of time? Because I just don't need any."

Ross Dawson: Things that some leaders will do. But, I mean, it comes back to the theme of amplifying cognition. The sense of — the real nub of the question is, yes, you can sort of say, "All right, well, now we are training the machine, and the machine gets better because it's interacting. We're giving it more work."

But it's really finding the ways in which the nature of the way we interact also increases the skills of the humans.

And so John Hagel talks about scalable learning. In fact, Peter Senge used to talk about organizational learning — and that's no different today. We have to be learning.

And so, saying, "Well, as we engage with the AI — and as you rightly point out — we are teaching and helping the AI to learn," we need to be able to build the process and systems and structures and workflows where the humans in it are not static and stagnant as they use AI more, but they’re more competent and more capable.

Kieran Gilmurray: Well, that's the thing we need to do, Ross.

Otherwise, what we end up with is something called cognitive offload — where now, all of a sudden, I'll get lazy, I'll let AI make all of the decisions, and over time, I will forget and not be valuable.

For me, this is a question of great potential with technology. But the real question comes down to: okay, how do we employ that technology?

And to your point a second ago — what do we do as human beings to learn the skills that we need to learn to be highly employable? To create, be more innovative, more creative using technology?

Ross Dawson: I answered the question you just asked.

Kieran Gilmurray: 100%, and this is — this is literally the piece here, so—

Ross: That's the question. So do you have any answers to that?

Kieran: No, of course. Of course. Well, mine is — it's that.

So, for me, AI will be — absolutely — and AI is massive. And let me explain that, because everybody thinks it's been around. If we look at generative AI for the last couple of years — but AI has been around for 80-plus years. It's what I call an 80-year-old overnight success story.

Everybody's getting excited about it. Remember, the excitement is down to the fact that I can now interact with — or you interact with — technology in a very natural sense and get answers that I previously couldn't.

So now, all of a sudden, we're experts in everything across the world. And if you use it on a daily basis, all of a sudden, our writing is better, our output's better, our social media is better.

So the first bit is: just learn how to use and how to interact with the technology.

Now, we mentioned a moment ago — but hold on a second here — what happens if everybody uses it all the time, the AI has been trained, there's a whole host of new skills?

Well, what will I do?

Well, this for me has always been the case. Technology has always come. There's a lot less saddlers than there are software engineers. There might be a lot less software engineers in the future.

So therefore, what do we do?

Well, my one is this. All of this has been the same, regardless of the technology: let technology do the bits that technology is really good at. Offload to it.

You still need to understand or develop your digital, your AI, your automation, your data literacy skills — without a doubt. You might do a little bit of offloading, because now we don't actually think about scientific calculators. We get on with it.

We don't go into Amazon and automatically work out all of our product sets, because it's got a recommendation engine. So therefore, let it keep doing all its stuff.

Whereas, as humans, I want to develop greater curiosity. I want to develop what I would describe as greater cognitive flexibility. I want to use the technology — now that I've got this — how can I produce even better, greater outputs, outcomes, better quality work, more innovative work?

And part of that is now going, "Okay, let the technology do all of its stuff. Free up tons of hours," because what used to take me weeks takes me days.

Now I can do other stuff, like wider reading. I can partner with more organizations. I can attempt to do more things in the day — whereas in the past, I was just too busy trying to get the day job done.

The other bits I would be saying: companies need to develop emotional intelligence in people.

Because now, if I can get the technology to do the stuff, now I need to engage with tech. But more importantly, I'm now freed up to work across silos, to work across businesses, to bring in different partner organizations.

And statistically, only 36% of us are actually emotionally intelligent.

Now, AI is an answer for that as well — but emotional intelligence should be something I would be developing inside of an organization. A continuous innovation mindset. And I'd be teaching people how to communicate even better.

Notice I'm letting the tech do all the stuff that tech should do regardless. Now I'm just over-indexing and over-amplifying the human skills that we should have developed over the last 10, 15, or 20 years.

Ross Dawson: Yeah. And so, your point — this comes about people working together. And so I think that was one of the — certainly one of the interesting parts of your book is around team dynamics.

So there's a sense of, yes, we have agentic systems. This starts to change the nature of workflows. Workflows involve multiple people. They involve AI agents as well.

So as we are thinking about teams — as in multiple humans assisted by technology — what are the things which we need to put in place for effective team dynamics and teamwork?

Kieran Gilmurray: Yeah, so — so look, what you will see potentially moving forward is that mixture of agentic labor working with human labor.

And therefore, from a leadership perspective, we need people — we need to teach people — to lead in new ways. Like, how do I apply agentic labor and human labor? And what proportion? What bits do I get agentic labor to do? What bits do I get human labor to do?

Again, we can't hand everything over to technology. When is it that I step in? Where do I apply humans in the loop?

When you look at agentic labor, it's going to be able to do things 24/7, but as people, we physically and humanly can't. So, how — when am I going to work? What is the task that I'm going to perform?

As a leadership or as a business — well, what are the KPIs that I'm going to measure myself on, and my team on? Because now, all of a sudden, my outputs potentially could be greater, or I'm asking people to do different roles than they've done in the past, because we can get agentic labor to do it.

So there's a whole host of what I would describe as current management consideration. Because, let's be honest — like when we introduced ERP, CRM, factory automation, or something else — it just changed the nature of the tasks that we perform.

So this is thinking through: where is the technology going to be used? Where should we not use it? Where should we put people? How am I going to manage it? How am I going to lead it? How am I going to measure it?

These are just the latest questions that we need to answer inside of work.

And again, from a skillset perspective — from both a leadership and getting my human labor team to do particular work, or how I onboard them — how do I develop them? What are the skills that I'm now looking for when I'm doing recruitment?

What are the career paths that I'm going to put in place, now that we've got human plus agentic labor working together?

Those are all conversations that managers, leaders, and team leaders need to have — and strategists need to have — inside of businesses.

But it shouldn't worry businesses, because again, we've had this same conversation for the last five decades. It's just been different technology at different times, where we had to suddenly reinvent what we do, how we do it, how we measure it, and how we manage it.

Ross Dawson: So what are specifics of how teams, team dynamics might work in using agentic AI in a particular industry or in a particular situation? Or any examples? So let's ground this.

Kieran Gilmurray: Yeah, so let's — let me ground it in physical robots before I come into software robots, because this is what this is: software labor, not anything else.

When you look at how factories have evolved over the years — so take Cadbury’s factory in the UK. At one stage, Cadbury's had thousands and thousands of workers, and everybody ended up engaging on a very human level — managing people, conversations every day, orchestration, organization. All of the division of labor stuff happened.

Now, when you go into Cadbury's factory, it's hugely automated — like other factories around the world. So now we're having to teach people almost to mind the robots.

Now we have far less people inside of our organizations. And hopefully — to God — this won't happen in what I’d describe as a knowledge worker park, but we're going to teach people how to build logical, organized, sequential things. Because to break something down into a process to build a machine — it's the same thing when it comes to software labor.

How am I going to break it and deconstruct a process down into something else? So the mindset needed to actually put software labor into place varies compared to anything else that we've done.

Humans were messy. Robots can’t be. They have to be very logical pieces.

In the past, we were used to dealing with each other. Now I'm going to have to communicate with a robot. That's a very different conversation. It's non-human. It's silicon — not carbon.

So how do I engage with a robot? Am I going to be very polite? And I see a lot of people saying, "Please, would you mind doing the following?" No — it’s a damn robot. Just tell it what to do. My mindset needs to change.

So if I take, in the past, when I'm asking someone to do something, I might say, "Give me three things" or "Can you give me three ideas?" Now, I've got an exponential technology where my expectations and requests of agentic labor are going to vary.

But I need to remember — I’m asking a human one thing and a bot another.

Let me give you an example. I might say to you, "Ross, give me three examples of..." Well, that’s not the mindset we need to adopt when it comes to generative AI. I should be going, "Give me 15, 50, 5,000," because it's a limitless vat of knowledge that we're asking for.

And then I need to practice and build human judgment — to say, "Actually, I'm not going to cognitively offload and let it think for me and just accept all the answers." But I'm now going to have to work with this technology and other people to develop that curiosity, develop that challenging mindset, to suddenly teach people how to do deeper research, to fact-check everything that I'm being told.

To understand when I should use a particular piece of information that’s been given to me — and hope to God it's not biased, not hallucinated, or anything else — but it’s actually a valuable knowledge item that I should be putting into workflow or a project or a particular document or something else.

So again, it's just working through: what is technology? What's the technology in front of me? What's it really good at? Where can I apply it?

And understanding that — where should I put my people, and how should I manage both?

What are the skills that I need to teach my people — and myself — to allow me to deal with all of this potentially fantastic, infinite amount of knowledge and activity that will hopefully autonomously deliver all the outcomes that I've ever wanted?

But not unfettered. And not left to its own devices — ever.

Otherwise, we have handed over human agency and team agency — and that's not something or somewhere we should ever go. The day we hand everything to the robots, we might as well just go to the care home and give up.

Ross Dawson: We'll be doing that soon. So around now, let's think about leadership.

So, I mean, you've alluded to that in quite a few — I mean, a lot of it has been really talking about some of the questions or the issues or the challenges that leaders at all levels need to engage with. But this changes, in a way, the nature of leadership.

As you say, you've got digital labor as well as human labor. The organization has a different structure. It impacts the boundaries of organizations and the flows of information and processes — cross-organizational boundaries.

So what is the shift for leaders? And in particular, what are the things that leaders can do to develop their capabilities for a somewhat different world?

Kieran Gilmurray: Yeah, it's interesting.

So I think there'll be a couple of different worlds here. Number one is, we will do what we've always done, which is: we'll put in a bit of agentic labor, and we'll put in a bit of generative AI, and we'll basically tweak how we actually operate. We'll just make ourselves marginally more efficient.

Because anything else could involve the redesign and the restructure of the organization, which could involve the restructure and the redesign of our roles. And as humans, we are very often very change-resistant.

Therefore, I don't mind technology that I understand, and I don't mind technology that makes me more productive, more creative. But I do mind technology that could actually disrupt how I lead, where I actually fit inside of the organization, and something else.

So for those leaders, there's going to be a minimal amount of change — and there's nothing wrong with that. That's what I call the "taker philosophy," because you go: taker, maker, shaper — and I'll walk through those in a second — which is, I'll just take another great technology and I'll be more productive, more creative, more innovative.

And I recommend every business does that at this moment in time. Who wouldn't want to be happier with technology doing greater things for you?

So go — box number one.

And therefore, the skills I'm going to have to learn — not a lot of difference. Just new skills around AI. In other words, understanding bias, hallucinations, understanding cognitive offloading, understanding where to apply the technology and not.

And by "not," I mean: very often people put technology at something that has no economic value. Waste time, waste money, waste energy, get staff frustrated — something else. So those are just skills people have to learn. It could be any technology, I've said.

The other method of doing this is almost what I describe as the COVID method. I need to explain that statement.

When COVID came about, we all worked seamlessly. It didn’t matter. There were no boundaries inside of organizations. Our mission was to keep our customers happy. And therefore, it didn’t matter about the usual politics, the usual silos, or something else. We made things work, and we made things work fast.

What I would love to see organizations doing — and very few do it — is redesign and re-disrupt how they actually work.

And I'm sitting there going, it's not that I'm doing what I'm doing and I've now got a technology — "Where do I add it on?" — as in two plus one is equal to three.

What I'm sitting going and saying is: How can I fundamentally reshape how I deliver value as an organization?

And working back from the customer — who will pay a premium for this — and therefore, if I work back from the customer, how do I reconstruct my entire business in terms of leadership, in terms of people, in terms of agentic and human labor, in terms of open ecosystems and partnerships and everything else — to deliver in a way that excites and delights?

If we take the difference between bookstore and Amazon — I never, or rarely, go into a bookstore anymore. I now buy Amazon almost every time, not even thinking about it.

If I look at AI-native labor — they’re what I describe as Uber’s children. Their experiences of the world and how they consume are very different than what you and I have constructed.

Therefore, how do I create what you might call AI-native intelligent businesses that deliver in a way that is frictionless and intelligent?

And that means: intelligent processes, intelligent people, using intelligent technology, intelligent leadership — forgetting about silos and breakdowns and everything else that exists politically inside of organizations — but applying the best technology. Be it agentics, be it automation, be it digital, be it CRM, ERP — it doesn’t really matter what it is.

Having worked back from the customer, design an organization to deliver on its promise to customers — to gain a competitive advantage.

And those competitive advantages will be less and less. I can copy all the technology quicker. Therefore, my business strategy won’t be 10 years. It possibly won’t be five. It might be three — or even less.

But my winning as a business will be my ability to construct great teams. And those great teams will be great people plus great technology — to allow me to deliver something digitally and intelligently to consumers who want to pay a premium for as long as that advantage lasts.

And it might be six months. It might be twelve months. It might be eighteen months.

So now we're getting to a phase of almost fast technology — just like we have fast fashion.

But the one thing we don’t want to do is play loose and fast with our teams. Because ultimately, I still come back to the core of the argument — that great people who are emotionally intelligent, who’ve been trained to question everything that they’ve got, who are curious, who enjoy working as part of a team in a culture — and that piece needs to be taken care of as well.

Because if you just throw robots at everything and leave very few people, then what culture are you actually trying to deliver for your staff and for your customers?

How do I get all of this work to deliver in a way that is effective, is affordable, is operationally efficient, profitable — but with great people at the core, who want to continue being curious, creating new and better ways of delivering in a better organization?

Not just in the short term — because we’re very short-termist — but how do I create a great organization that endures over the next five or ten years?

By creating flexible labor and flexible mindsets, with flexible leaders organizing and orchestrating all this — to allow me to be a successful business.

Change is happening too quickly these days. Change is going to get quicker.

Therefore, how do I develop an adaptive mindset, adaptive labor force, and adaptive organization that's going to survive six months, twelve months — and maybe, hopefully to God, sixteen months plus?

Ross Dawson: Fantastic. That's a great way to round out. So where can people find out more about your work?

Kieran Gilmurray: Yeah, look, I'm on LinkedIn all the time — probably too much. I should get an agentic labor force to sort that out for me, but I'd much prefer authentic relationships than anything else.

Find me on LinkedIn — Kieran Gilmurray. I think there are only two of me: one’s in Scotland, who is related some way back, and the Irish one.

Or www.kierangilmurray.com is where I publish far too much stuff and give far too much stuff — things — away for free. But I have a philosophy that says all boats rise in a floating tide. So the more we share, the more we give away, the more we benefit each other.

So that's going to continue for quite some time.

I have a book out on agentic AI. Again, it's being given away for free. Ross, if you want to share it, please go for it, sir, as well.

As I said, let's continue this conversation — but let's continue this conversation in a way that isn't about replacing people. It's about great leadership, great people, and great businesses that have people at their core, with technology serving us — not us serving the technology.

Ross: Fabulous. Thanks so much, Kieran.

Kieran: My pleasure. Thanks for the invite.

The post Kieran Gilmurray on agentic AI, software labor, restructuring roles, and AI native intelligence businesses (AC Ep84) appeared first on Humans + AI.

"We humans often tend to be very restricted—even when we are world champions in a game. And I'm very optimistic that AI will surprise us, with very different ways of solving complex problems—and we can make use of that."

– Jennifer Haase

Dr. Jennifer Haase is a researcher at the Weizenbaum Institute, and lecturer at Humboldt University and University of the Arts Berlin. Her work focuses on the intersection of creativity, Artificial Intelligence, and automation, including AI for enhancing creative processes. She was named as one the 100 most important minds in Berlin science.

Website:

Jennifer Haase

Jennifer Haase

LinkedIn Profile:

Jennifer Haase

Websites & Platforms

jenniferhaase.com

ChatGPT

Concepts & Technical Terms

Artificial Intelligence (AI)

Human-AI Co-Creativity

Generative AI

Large Language Models (LLMs)

ChatGPT

GPT-4

GPT-3.5

GPT-4.5

Business Informatics

Psychology

Creativity

Divergent Thinking

Convergent Thinking

Mental Flexibility

Iterative Process

Everyday Creativity

Alternative Uses Test

Creativity Measures

Creative Performance

Ross Dawson: Jennifer, it's a delight to have you on the show.

Jennifer Haase: Thanks for inviting me.

Ross: So you are diving deep, deep, deep into AI and human co-creativity. So just to hear—just back a little bit—sort of how you've embarked on this journey. I mean, love to—we can fill in more about what you're doing now. But how did you come to be on this journey?

Jennifer: I would say overall, it was me stumbling into tech more and more and more. So I started with creativity.

My background is in psychology, and I learned about the concept of creativity in my Bachelor studies, and I got so confused, because what I was taught was nothing like what I thought creativity was—or how it felt to me.

It took me years to understand that there are a bunch of different theories, and it was just one that we were taught. But that was the spark of the curiosity for me to try to understand this concept of creativity. And I did it for years.

Then, by pure luck, I started a PhD in Business Informatics, which is somewhat technical. The lens of how I looked at creativity shifted from the psychological perspective more into the technical realm, and I looked at business processes and how they are advanced by general technology—basic software, basically.

Then I morphed—also, by sheer luck—I morphed into computer science from a research perspective. And that coincided with ChatGPT coming around, and this huge LLM boom happened two, three years ago.

And since then, I'm deeply in there. I just fell, fell in this rabbit hole.

Ross: Yeah, well, it's one of the most marvelous things. So the very first use case for most people, when they first use ChatGPT, is: write a poem in the style of whatever, or essentially creative tasks. And pretty decently does those to start off—until you sort of started to see the limitations at the time.

Jennifer: Yeah, and I think it did so much. It's so many different perspectives.

I think we—as I said, I studied creativity for quite a while—but it was never as big of a deal, let's say. It was just one concept of many. But since AI came around, I think it really threatened, to some part, what we understood about creativity, because it was always thought of as this pinnacle of humanness—right next to ethics.

And I think intelligence had its bumps two or three decades ago, but for creativity, it was rather new. So the debate started of what it really means to be creative.

I think a lot of people also try to make it even bigger than it is. But I think it is as simple as—a lot about creativity is, for example, in terms of poets—poetry is language understanding, right? And so LLMs are really good at it. And it's just the case. It's fine.

I think we can still live happy lives as humans, although technology takes a lot over.

Ross: Yes. So humans are creative in all sorts of dimensions. AI has complementary—let's say, also different—capabilities in creativity.

And in some of your research, you have pointed to different levels of how AI is supporting us in various guises—through being a tool and assistant, through to what you described as the co-creation. So what does that look like?

What are some of the manifestations of human-AI co-creativity, which implies peers with different, complementary capabilities?

Jennifer: Yeah, I think the easiest way to look at it is if you imagine working creatively with another person who is really competent—but the person is a technical version of it, and usually we call that AI, right? Or generative AI these days.

So the idea is that you can work with a technical tool from an eye-to-eye level. Really, the tool would have a—well, now we're getting into the realm of using psychological terms, right—but the tool would have a decent enough understanding so it would appear competent in the field that you want to create.

I think the biggest difference we see to most common tools that we have right now—which I would argue are not on this level yet—tools like ChatGPT and others, they follow your lead, right? If you type in something, they will answer, sometimes more or less creatively.

But you can take that as inspiration for your own creativity and your own creative process. And that really holds big potential. It's great.

But what we are envisioning—and seeing in some parts already happening in research—I think this is the direction we're going to and really want to achieve more: that we have tools that can also come up with ideas, or important input for the creative problem.

Not—when I say on their own—I don't mean that they are, I don't know, entities that just do. But they contribute a significant, or really a significant part of the creative process.

Ross: So, I mean, we'll come back a little bit to the distinctions between how AI creativity contrasts to human creativity. But just thinking about this co-creative process—from your research or other research that you're aware of—what are the success factors? What are the things which mean that that co-creation process is more likely to be fruitful than not?

Jennifer: I think it starts really with competence. And I think this is something, in general, we see that generative AI just became extremely good at, right?

They know, so to speak, a lot and tailor a lot of knowledge, and that is very, very helpful—because we need broad associations, coming from mostly different fields, and connect that to come up with something we consider new enough to call it creative.

That is a benefit that is beyond human capabilities, right? What we see right now those tools are doing—that is one part. But that is not all.

What you also need is the spark of: why would something need to be connected? And I think that is especially where raising the creative questions, coming up with the goal that you want to achieve something too, is still the human part.

But—it doesn’t need to be. That’s all I’m saying. But still, it is.

Ross: So, I mean, there are some—very crude workflows, as in, you get AI to ideate, then humans select from those, and then they add other ideas, or you get humans and then AI sort of combines, recombines.

Are there any particular sequences or flows that seem to be more effective?

Jennifer: It's interesting. I think this is also an interesting question for human creative work alone, even without technology—like, how do you achieve the good stuff, right?

And I think what you just described, for me, would be kind of like a traditional way of: oh, I have a need, or I have a want—like, I want to create something, or I want to solve something, or I need a solution for a certain problem. And I describe that, and I iterate a best solution, right?

This is part of what we call the divergent thinking process. And then, at a certain point, you choose a specific solution—so you converge.

But I think where we have mostly the more interesting creative output—for humans and now also especially with AI—is that you kind of reverse the process. So let's assume you have a solution and you need to find issues for it.

For example, you have an invention. I think—yeah, I think it was that there's this story told about the Post-its, you know, the yellow Post-its. So they were kind of invented because someone came up with glue that does not stick at all—like, really bad glue.

And they had this as the final product. Now it's like, "Okay, where can you make use of it?" And then they came up with, "Oh, maybe, if you put it on paper, you can come up with these sticky notes that just glue enough." So they hold on surfaces, but they don't stick forever, so you can easily erase them.

They're very practical in our brainstorming work, for example.

And this kind of reverse thinking process—it's much more random. And for many people, it's much more difficult to open up to all the possibilities that can be.

What I've seen is that if you try to poke LLMs with such very diverse, open questions, it can be very interesting what kind of comes out there.

Ross: Though, to your point, I mean, this is the way—the human frames, the AI can respond. But the human needs to frame—as in, “Here is a solution. What are ways to be able to apply?”

Jennifer: And all the examples—like, what I'm thinking of right now—is what is working with the tools that we have with LLMs.

And I think what you were asking me before about the fourth level that we described with this co-creation—these are tools that work a bit differently. These are tools that, for now, mostly exist in research because you still need a high level of computational knowledge.

So, the work that I did—the colleagues that I work with—are from computer science or mathematicians who program tools that know some rules of the game, or some—let's call them—boundary conditions of our creative problem that we are dealing with.

And then the magic—or the black box magic—of AI is happening. And something comes out. And sometimes we don’t really understand what was going on there. We just see the results.

And then, with such results, we can iterate. Or maybe something goes in the direction as we assume could be part of the solution.

So it becomes this iterative process between an LLM or AI tool doing something, we’re seeing the results, saying yes or no, nudging it into different directions, and so, overall, coming up with a potentially proper solution.

This is—at least in the examples that we see.

And if you have such a process and look over it, like what was happening, often what we see is that LLMs or AI tools in general—with their, let’s call it, broad knowledge, or the very intense, broad computational capacities that they have—they do stuff differently than we as humans tend to do stuff.

And this is where it becomes interesting, right? Because now we are not bounded in this common way of thinking and finding associations, or iterating smaller solutions.

Now we have this interesting artificial entity that finds very different ways of solving complex problems—and we can make use of that.

Of course, we can learn from that.

Ross: Absolutely. And I think you've pointed to some examples in your papers. I mean—other, sort of, I suppose we've been quite conceptual—so examples that you can give of either what people have done, or projects you've been involved with, or just types of challenges?

Jennifer: I think—to explain the mechanism that I'm talking about—I think the first creative, artificial example, like the real, considered properly creative example, was when AlphaGo, the program developed to play Go—the game similar to, or somewhat similar to, chess but not chess—when this tool was able to come up with moves, like play moves, which were very uncommon.

Still within the realm of possibilities, but very, very uncommon to how humans used to play.

And so, I think what this new was back in 2016, right? When this happened—when DeepMind, from Google, built this tool and kind of revolutionized AI research.

What it showed us is exactly this mechanism of these tools. Although they are still within the realm of possibilities—still within what we consider the rules, right, of the game—it showed some moves which were totally uncommon and surprising.

And I think this shows us that we humans often tend to be very restricted. Even when we are world champions in a game, we are still restricted to what we commonly do—what is considered a good rule of thumb for success.

And I'm very optimistic that AI will surprise us, like in this direction—with this mechanism—quite a lot in the future.

Ross: Yeah, and certainly, related to what you're describing, some similar algorithms have been applied to drug discovery and so on.

Part of it is the number-crunching, machine learning piece, but part of it is also being able to find novel ways of folding proteins or other combinations which humans might not have envisaged.

Jennifer: Yeah, exactly. And exactly—it's in part because these machines are just so much more advanced in how much, or how many, information they can hold and combine.

This is, in part, purely computational. It's a bit unfair to compare that to our limited brains. But it's not just that. It's not just pure information, right?

It's also how this information is worked upon, or the processes—how information is combined, etc. So I think there are different levels of how these machines can advance our thinking.

Ross: So one of the themes you've written about is designing for synergies—how we can design so that we are able to be complementary, as opposed to just delegating or substituting with AI.

So what are those design factors, or design patterns, or mentalities we need?

Jennifer: Well, I will propose, first up—I think it's extremely complicated. Not complicated, but it will become a huge issue.

Because, let's say, if technology becomes so good—and we see that right now already with LLMs like ChatGPT—it’s so easy for us. And I mean that in a very neutral way. But lazy humans as we are—I think we are inherently lazy—it’s really tough for us to keep motivated to think on our own, to some degree at least, and not have all the processes overtaken by AI.

So, saying that, I think the most essential, most important part whenever we are working with LLMs is: we have to keep our motivation in the loop—and our thinking to some degree in the loop—within the process.

And so, we need a design which engages us as humans.

I think it's easily seen right now with LLMs. When you need the first step in—like typing some kind of prompt, or even in a conversation—you have to initiate it, right? You have to come up with, maybe even, your creative task at first.

And I think this will always be true, because we humans control technology by developing it, right?

But even when you're more on the user end—forcing us to be in the loop, and thinking it through, and controlling the output, etc.—is one part.

But I think what it also needs, especially for the synergy, is for the technology to adapt to us—to serve us, so to speak.

And I think this is an aspect that is a little bit underdeveloped right now. What do I mean by that?

I want a tool that serves me in my thinking. It should be competent enough that I perceive it as a buddy—eye to eye. That is the vision that I have.

But I still always want the control. And I want it to adapt to me, and that I don't have to adapt too much to the tool.

Right now, we’re mostly just provided with tools that we need to learn how to deal with. We need to understand how prompting works, etc., etc. And I want that reversed.

I want tools which are competent enough to understand, “Okay, this is Jenny. She is socialized in this way. She usually speaks German,”—whatever kind of information would be important to get me involved and understand me better.

I think this is the vision for synergy that I'm thinking of.

Ross: No, I really like that. The idea of designing for engagement, because instead of saying, yeah, why is it going to make us want to be engaged and continue the process and want to want to be involved, as opposed to doing the hard work of telling the—keep on telling the AI to do stuff.

Jennifer: Yes, and also sometimes—I mean, I work a lot with ChatGPT and other similar tools—and sometimes I'm like, I found myself, I hope I don't spoil too much, but sometimes I find myself copy-pasting too much because there's nothing left for me to do.

And to some degree, it can happen that the tools are too good, right? Because they are meant to create the output as the output, but they are not meant to be part of this iterative thinking process.

I think you can design it much better and easier to go hand in hand with what I'm thinking and what I want to advance. Maybe.

Ross: Yeah, yes, otherwise the onus is on the human to do it all. So in one of your papers, you identify—you used a number of the different models, and I believe you found that GPT-4 was the best for a variety of ideation tasks.

But you've also done some more recent research. I'd love to hear about strengths, weaknesses, or different domains in which the different models are good, or—

Jennifer: Yeah, that's quite interesting, right? Because—okay, so going back to the start of the big—let's call it the big boom of LLMs, right?

I think it was early '23, right, when ChatGPT came around. End of '22. Okay, so it took a while when it reached Germany—it was for us. No, just joking.

But okay, so around this time, what we found was intense debates arguing that, although these tools are generative, they cannot be creative. And that was the stance held tightest—maybe especially from creativity researchers and mostly psychologists, right?

As I mentioned before, it's a little bit of this fear that too much is taken over by technology. I think that is a strong contributor—even among researchers.

So what we went out to do is—we basically wanted to ask LLMs the same creativity measures as we would do for humans. Like, when you want to know if a person holds potential for creative thinking, you ask them creative questions, and they have to perform—if they want to.

And that's exactly what we did with LLMs.

Back in the day, we did it with the LLMs that were easily reachable and free in the market—like ChatGPT. And now, we really redid it with the current LLMs, with the current versions.

And—I don't know if you've seen that—but most LLMs are advertised, when the new versions come out, usually they are advertised with: they are more competent, and they are more creative.

And so we questioned that. Is that really true? Is ChatGPT 4.5, for example—the current version—is it more creative than 3.5 back in the day?

And what we find is—it's so messy, actually. Because for some tools, yes, they are a bit more creative than they used to be two years ago. But the picture is really not clear.

You cannot really tell or say or argue that the current versions we are having are more creative than two years ago—or even more creative than humans.

It's been interesting. We're not really sure why. But all we can say is that, on average, these tools are as good at coming up with everyday-like uses or everyday-like ideas for everyday problems.

They are, on average, as good as humans—random humans picked from surveys.

And I think that is good news, right? Because LLMs are easier to ask than random humans most of the time.

But the promise that they become more and more creative with every new release, in our perspective, does not hold up.

So that is the bigger, bigger picture. Let's start there.

Ross: So that’s very interesting. So this is using some of the classic psychological creativity tests. And so you're applying what has for a long time been used for assessing creativity in humans, and simply applying exactly the same test to LLMs?

Jennifer: And to be fair, within the creativity research community, we agree that those tests are not good. Okay, they're really pragmatic. We totally agree on that, so we do not have to fight for this point.

But it's commonly what we use to assess human potential for creative thinking—or even more concise, for divergent thinking—which is only one important, but just one aspect, of the whole creative journey, let's say.

And it basically just asks how good you are, on the spot, at coming up with alternative uses for everyday products like a shoe or toothbrush or newspaper.

And of course, you can come up with obvious uses. But then there are the creative ones, which are not so easy to think of, right? And LLMs are good at that.

They will deliver a lot of ideas, and quite a few of those are considered original compared to human answers.

We also now used another test, which is a little bit more arbitrary even, but it proved to be somewhat of a good predictor for creative performance overall. And that is: you are asked to come up with 10 words which are as different from each other as possible.

So very pragmatic again.

And these LLMs—as they, you know, know one thing, and that is language—are, again, quite good at that on average.

But it's not that you see that they are above average, or that a specific LLM would be above average. We see some variety, but the picture, I would say, is not too clear.

And also, to mention—which was a little bit surprising to us, actually—is that those LLMs, we asked them several times, like, a lot of times, and the variance in terms of originality—the variance is quite huge.

So if you ask an LLM like ChatGPT for creative ideas, sometimes you can have quite a creative output, and sometimes it's just average.

Ross: So you did say that you're comparing them to random humans. So does that mean that generally perceived-to-be-creative humans are significantly outperforming the LLMs on these tasks?

Jennifer: Yeah, yeah. So, but the thing is, there is usually no creative human per se. So there's nothing about a human that makes a human per se creative.

We tend to differ a little bit on how well we perform on such tasks. Yes, we do differ in our mental flexibility, let's say. But a creative individual is usually an individual which found a very good fit between their thinking, their experience, and the kind of creative task they're doing.

And just think about it, because this creativity can be found in all sorts of domains, right? And people can be good or less good in those domains, and that correlates highly with the creativity.

So when we ask about the general, like, the ideas for everyday tasks, there is not really the creative individual, right?

They are motivated individuals, which makes a huge difference for creativity measures. If you're motivated and engaged, that is something we take as granted.

For LLMs, I guess if you compare them, the motivation is there.

But what we see in terms of the best answers—the most original answers in our data sets—most of the time, not all, but most of the time, come from humans.

Ross: Very interesting. So, this is the Amplifying Cognition podcast, so I want to sort of round up by asking: all right, so what's the state of the nation or state of the world, and where we are moving in terms of being able to amplify and augment human cognition, human creativity?

So I suppose that could be either just, improving human creativity, or collaborating, or, you know, this co-creativity.

Jennifer: I think the potential for significant improvements and amplifications has never been better. But I think at the same time as I'm saying that, I think the risks have never been higher.

And that is because, as I said, we are lazy people. That's just what humanist means—and that is fine—but it also means that we have a great risk of using these technologies not for us, but being used by them, basically, right?

So we can use ChatGPT and other tools to do the task for us, or we can use them to do the task more efficiently and better with them.

I think this difference can be very gradual, very minor, but it makes the whole difference between success and big dependencies—and potentially failure.

Ross: Yeah, and I think you make a point—which I often also do—which is over-reliance is the biggest risk of all, potentially.

Where, if we start to just sort of say, “This is good, I'll let the AI do the task, or the creativity, or whatever,” it's dangerous on so many levels.

Jennifer: Because it does good enough most of the time, right?

Technology became so good for many tasks—not all, but many tasks—that it does it good enough. And I think that is exactly where we have the potential to become so much better, right?

Because if you now take the time and effort that we usually would put into the task itself, we could just improve on all levels.

And that is the potential I'm talking about. I think a lot is to be advanced, and a lot is to be gained—if we play it right.

Ross: And so, what's on your personal research agenda now?

Jennifer: Oh, I fell into this agentic LLM hole.

Yeah, no, no—it's not just looking at individual LLMs, but to chain them and combine them into bigger, more complex systems to have—or work on—bigger and complex issues, mostly creative problems, and see where the thinking of me and the tool, yeah, excels, basically, right?

And where do I, as a human, have to step in to fine-tune specific bits and pieces and really find the limits of this technology if you scale it up?

That's my agenda right now.

Ross: I'm very much looking forward to reading the research as you publish it. 

Jennifer: Thank you. 

Ross: Is there anywhere people can go to find out more about your work?

Jennifer: Yeah, I collect everything on jenniferhaase.com. That's my web page. It's hugely up to date there, and you can find talks and papers.

Ross: Fabulous. Love the work you're doing. Jennifer, thanks so much for being on the show and sharing.

Jennifer: Thank you very much. It was—yeah, I love to talk about that, so thanks for inviting me.

The post Jennifer Haase on human-AI co-creativity, uncommon ideas, creative synergy, and humans outperforming (AC Ep83) appeared first on Humans + AI.

“We should not make technology so that we can be stupid. We should make technology so we can be even smarter… not just make the machine more intelligent, but enhance the overall intelligence—especially human intelligence.”

–Pat Pataranutaporn

Pat Pataranutaporn is Co-Director of MIT Media Lab’s new Advancing Humans with AI (AHA) research program, alongside Pattie Maes. In addition to extensive academic publications, his research has been featured in Scientific American, MIT Tech Review, Washington Post, Wall Street Journal, and other leading publications. His work has been named in TIME’s “Best Inventions” lists and Fast Company’s “World Changing Ideas.”

Websites:

MIT Media Lab

AI (AHA)

LinkedIn Profile:

Pat Pataranutaporn

Reimagining ai as a tool for human flourishing

Exploring the future you project and long-term thinking

Boosting motivation through personalized ai learning

Enhancing critical thinking with question-based ai prompts

Designing agents that collaborate, not dominate

Preventing collective intelligence from becoming uniform

Launching aha to measure ai’s real impact on people

People

Hal Herschfeld

Pattie Maes

Elon Musk

Organizations & Institutions

MIT Media Lab

KBTG

ACM SIGCHI

Center for Collective Intelligence

Technical Terms & Concepts

Human flourishing

Human-AI interaction

Digital twin

Augmented reasoning

Multi-agent systems

Collective intelligence

AI bias

Socratic questioning

Cognitive load

Human general intelligence (HGI)

Artificial general intelligence (AGI)

Ross Dawson: Pat, it is wonderful to have you on the show.

Pat Pataranutaporn: Thank you so much. It's awesome to be here. Thanks for having me.

Ross: There’s so much to dive into, but as a starting point: you focus on human flourishing with AI, exactly. So what does that mean? Paint the big picture of AI and how it can help us to flourish as who we are and our humanity.

Pat: Yeah, that's a great question. So I'm a researcher at MIT Media Lab. I've been working on human-AI interaction before it was cool—before ChatGPT took off, right?

So we have been asking this question for a long time: when we focus on artificial intelligence, what does it mean for people? What does it mean for humanity?

I think today, a lot of conversation is about how we can make models better, how we can make technology smarter and smarter. But does that mean that we can be stupid? Does it mean that we can just let the machine be the smart one and let it take over?

That is not the vision that we have at MIT. We believe that technology should make humans better.

So I think the idea of human flourishing is an umbrella term that we use to describe different areas where we think AI could enhance the human experience.

For me in particular, I focus on three areas: how AI can enhance human wisdom, enhancing wonder, and well-being. So: 3 W’s—wisdom, wonder, and well-being.

We work on many projects to look into these areas. For example, how AI could allow a person to talk to their future self, so that they can think in the longer term, to see that future more vividly. That’s about enhancing wonder and wisdom.

We think a lot about how AI can help people think more critically and analyze information that they encounter on a daily basis in a more comprehensive way.

And you know well-being, we have many projects that look at how AI can improve human mental health, positive thinking, and things like that.

But at the end, we also focus on AI that doesn’t lead to human flourishing, to balance it out. We study in what contexts human-AI interaction leads to negative outcomes—like people becoming lonelier or experiencing negative outcomes such as false memories, misinformation, and things like that.

As scientists, we’re not overly optimistic or pessimistic. We're trying to understand what's going on and how we can design a better future for everyone. That's what we're trying to focus on. Yeah?

Ros: Fabulous. And as you say, there are many, many different projects and domains of research which you're delving into. So I'd like to start to dive into some of those.

One that you mentioned was the Future You project. So I'd love to hear about what that is, how you created it, and what the impact was on people being able to interact with their future selves.

Pat: Totally. So, I mean, as I said, right, the idea of human flourishing is really exciting for us. And in order to flourish, like, you cannot think short term. You need to think long term and be able to sort of imagine: how would you get there, right?

So as a kid, I was interested in sort of a time machine. Like, I loved dinosaurs. I wanted to go back into the past and also go into the future, see what would happen in the future, like the exciting future we might have. So I really love this idea of, like, having a time machine.

And of course, we cannot do a real time machine yet, but we can make a simulation of a time machine that uses a person’s personal data and can extrapolate that, and use other data to kind of see, okay, if the person has this current behavior, things that they care about, what would happen down the road—like what would happen in the future.

So we built an AI simulation that is a digital twin of a person. And we first ask people to kind of provide us with some basic information: their aspiration, things that they want to achieve in the future. And then we use the current behavior that they have to kind of create what we call a synthetic memory, or a memory that that person might have in the future, right?

So normally, memory is something that you already experienced. But in this case, because we want to simulate the future self, we need to build memory that you did not experience yet but might actually experience in the future.

So we use language model combined with the information that the person gives us to create this sort of intermediary representation of person experience, and then feed that into a model that then allows us to create human-like conversation.

And then we also age the image of the person. So when the person uploads the image, we also use a visual model that can kind of create an older representation of that person. And then combine these together, we are creating an AI-simulated future self that people can have a conversation with.

So we have been working with psychologists—Professor Hal Herschfeld from UCLA—who looks at the concept of future self-continuity, which is a psychological concept that measures how well a person can vividly imagine their future self. And he has shown that if you can increase this future self-continuity, people tend to have better mental health, better financial saving, better decision, because they can kind of think for the long term, right?

So we did this experiment where we created this future self system and then tested it with people and compared it with a regular chatbot and having no intervention at all. And we have shown that this future self intervention can increase future self-continuity and also reduce people’s anxiety as well.

So they become much more of a future thinker—not only think about today’s situation, but can see the possibility of the future and have better mental health overall. So I think this is really exciting for us, because we built a new type of system, but also really showed that it had a positive impact in the real world.

Ross: What were the ranges of ages of people who were involved in this research?

Pat: Yeah, so right now, the prototype that we developed is for younger population—people that just finished college or people that just finished high school, people that still need to think about what their future might look like, people that still would benefit from having ability to kind of think in the longer term.

And right now, we actually have a public demo that everyone can use. So people can go to our website and then actually start to use it. You can also volunteer the data for research as well. So this is sort of in the wild, or in the real world study. That’s what we are doing right now.

So if people like to volunteer the data, then we can also use the data to kind of do future research on this topic. But right now, the system has been used by people in over 190 countries, and we are really excited for this research to be in the real world and have people using it.

Ross: Fabulous. We'll have the link in the show notes.

So, one of the other interesting aspects raised across your research is the potential positive impact of AI on motivation. I think that's a really interesting point. Because, classically, if you think about the future of education, AI can have custom learning pathways and so on. But the role of the human teachers, of course, is to inspire and to motivate and to engage and so on.

So I'd love to hear about how you're using AI to develop people's positive motivation.

Pat: Yeah, that's a really great question. And I totally agree with you that the role of the teacher is to inspire and create this sort of positive reinforcement or positive encouragement for the student, right? We are not trying to replace that.

Our research is trying to see what kind of tools the teacher can use to improve student motivation, right? And I think today, a lot of people have been asking, like, well, we have AI that can do so many things—why do we need to learn, right?

And we believe at MIT that learning is not just for the benefit of getting a job or for the benefit that you will have a good life, but it's good for personal growth, and it's also a fun process, right? Learning something allows you to feel excited about your life—like, oh, you can now do this, even though AI can do that.

I mean, a car can also go from one place to another place, but that doesn't mean we should stop walking, right? Or you can go to a restaurant and a professional chef can cook for you, but it's also a very fun thing to cook at home, right? With your loved ones or with your family, right?

So I think learning is a really important process of being human, and AI could make that process even more interesting and even more personal, right?

We really emphasize a lot on the idea of personalized learning, which means that learning can be tailored to each individual. People are very different, right? We learn in different ways. We care about different things.

And learning is also about connecting the dots—things that we already know and new things that we haven't learned before. How do we connect that dot better?

So we have built many AI systems that try to address these.

The first project we looked at was what happens if we can create virtual characters that can work with teachers to help students learn new materials. They can be a guest lecturer, they could be a virtual tutor that students can interact with in addition to their real teacher, right?

And we showed that by creating characters based on the people that students like and admire—like, at that time, I think people liked Elon Musk a lot (I don’t know about now; I think we would have a different story)—but at that time, Elon Musk was a hero to many people.

So we showed that if you learn from virtual Elon Musk, people have a higher level of learning motivation, and they want to learn more advanced material compared to a generic AI.

So personalization, in this case, really helped with enhancing personalized feeling and also learning motivation and positive learning experience. We have shown this across different educational measures.

Another project we did was looking at examples, right? When you learn things, you want examples to help you understand the concept, right? Sometimes concepts can be very abstract, but when you have examples, that’s when you can start to connect it with the real world.

Here we showed that if we use AI to create examples that resonate with the student’s interests—like if they love Harry Potter, or, I don’t know, like Kim Kardashian, or whatever—Minecraft or whatever things that people like these days, right? Well, I feel like an old person now, but yeah, things that people care about.

If you create an example using elements that people care about, we can also make the lesson more accessible and exciting for people as well, right?

So this is a way that AI could make learning more positive and more fun and engaging for students. Yeah.

Ross: So one of the domains you've looked at is augmented reasoning. And so I think it's a particularly interesting point now. In the last six months or so, we've all talked about reasoning models with large language models—or perhaps “reasoning” in quotation marks.

And there are also studies that have shown in various guises that people do seem to be reducing their cognitive engagement sometimes, whether they're overusing LLMs or using them in the wrong ways. So I'd love to hear about your research in how we can use AI to augment reasoning as well as critical thinking capabilities.

Pat: That's a great question. I mean, that’s going back to what I said, right? Like, what does it mean for humans to have smart models around us? Does it mean we can be stupid?

I think that’s a degradation of humans, right? We should not make technology so that we can be stupid. We should make technology so we can be even smarter, right?

So I think the end goal of having a machine or models that can do reasoning for us, rather than enhance our reasoning capability—I think that's the wrong goal, right? And again, if you have the wrong outcome or the wrong measurement, you're gonna get the wrong thing.

So first of all, you need to align the goal in the right direction.

That's why, in my PhD research, I really want to focus on things that ultimately have positive impact on people. AI models continue to advance, but sometimes humans don't advance with the AI models, right?

So in this case, reasoning is something that's very, very critical. You can trace it back to ancient Greek. Socrates talked a lot about the importance of questioning and asking the right question, and always using this critical thinking process—not trusting things at face value, right?

We have been working on systems—again, the outcome of human-AI interaction can be influenced by both human behavior and AI behavior, right? So we can design AI systems that engage people in critical thinking rather than doing the critical thinking for them. That could be very dangerous, right?

These systems right now don't really have real reasoning capability. They're doing simulated reasoning. And sometimes they get it right because, on the internet, people have already expressed reasoning and thinking processes. If you repeat that, you can get to the right answer.

I mean, the internet is bigger than we imagined. I think that's what the language models show us—that there's always something on the internet that allows you to get to the right answer. You have powerful models that can learn those patterns, right?

So these models are doing simulated reasoning, which means they don't have real understanding. Many people have shown that right now—that even though these systems perform very well on benchmarks, in the real world they still fail, especially with things that are very unique and very critical, right?

So in that case, the model, instead of doing the reasoning for us, could make us have better reasoning by teaching us the critical thinking process. And there are many processes for that. Many schools of thought.

We have looked at two processes. One of them is in a project called Variable Reasoner. We made a wearable device—like wearable smart glasses—with an AI agent that runs the process of verifying statements that people listen to and identify and flag when the statement people listen to has no evidence to support, right?

This is really, really important—especially if you love political speeches, or you love watching advertisements or TikTok. Because right now, social media is filled with statements that sound so convincing but have no evidence whatsoever.

So this type of system can help flag that. Because, as humans, we tend to go—or we tend to follow along—if things sound reasonable, sound correct, sound persuasive, we tend to go with them. But things that sound persuasive or sound correct doesn't mean it's correct, right?

It can use all sorts of heuristics and other fallacies to get you to fall into that trap. So our system—the AI—can be the system that follows things along and helps us flag that for us.

We have shown that when people wear these glasses, when the AI helps them think through the statements they listen to, people tend to agree more with statements that are well-reasoned and have evidence to support, right?

So we can show that we can nudge people to pay more attention to the evidence part of the information they encounter.

That's one project.

Another project—we borrowed the technique from Socrates, the ancient Greek philosopher. We showed that if the AI doesn’t give the answer to people right away but rather asks a question back—it’s kind of counterintuitive, like, well, but people need to arrive at that information for themselves—

We showed that when the AI asked questions, it improved people’s ability to discern true information from false information better than AI giving the correct answer.

Which some people might ask: why is that the case?

And I think it's because people already have the ability. Many of us already have the ability to discern information. We are just being distracted by other things.

So when the AI asks a question, it can help us focus on things that matter—especially if the AI frames the information in a way that makes us think, right?

For example, if there is a statement like: "Video games lead to people becoming more violent," and the evidence is "a gamer slapped another last week." For example—

If the AI starts to frame that into: "If one person stabs another person, does that mean that every gamer will become violent after playing video games?"

And then you start to realize that, oh, now there's an overgeneralization. You're using the example of one to overgeneralize to everyone, right?

If the AI frames the statement into a question like this, some people will be able to come up with the answer and discern for themselves. And this not only allows them to reach the right and correct answer but also strengthens their process as well, right?

It's kind of like AI creating or scaffolding our critical thinking so that our critical thinking muscle can be strengthened, right?

So I think this is a really important area of research. And there are many more research coming out that show how we can design AI systems that enhance critical thinking rather than doing the critical thinking for us.

Ross: So in a number of other domains, there's been research which has showed that whilst in some contexts AI can produce superior cognition or better thinking abilities, when the AI is withdrawn, they revert back.

So one of the things is not only using AI in the enhancement process, but post-AI—to actually enhance the norms. When you don't have the AI, that you're still able to enhance your critical thinking.

So has that been demonstrated, or is that something you would look at?

Pat: Yeah, that's a really important question. We haven't looked at a study in that sort of domain—what happens when people stop using the AI, or what happens when the AIs are being removed from people—but that's something that is part of the research roadmap that we are doing.

At MIT right now, there's a new research effort called AHA. We want to create aha moments, but AHA also stands for Advancing Humans with AI. And the emphasis is on advancing humans, right? AI is the part that's supposed to help humans advance. So the focus is on the humans.

We have looked at different research areas. We've already been doing a lot of work in this, but we are creating this roadmap for what future AI researchers need to focus on—and this is part of it.

This is the point that you just mentioned: the idea of looking at what happens when the AI is removed from the equation, or when people no longer have access to the technology. What happens to their cognitive process and their skills? That is a really important part that is part of our roadmap.

And so, for the audience out there—this April 10 is when we are launching this AHA research program at MIT. We have a symposium that everyone can watch. It's going to be streamed online on the MIT Media Lab website. You can go to aha.media.mit.edu, and see this symposium.

The theme of this symposium is: Can we design AI for human flourishing? And we have great speakers from OpenAI, Microsoft. We have great thinkers like Geraldine, Tristan Harris, Sherry Turkle, Arianna Huffington, and many amazing people who are joining us to really ask this question.

And hopefully, we hope that this kind of conversation will inspire the larger AI researchers and people in the industry to ask the important question of AI for human flourishing—not just AI for AI's sake, or AI for technological advancement's sake.

Ross: Yeah, I've just looked at the agenda and the speakers—this is mind-boggling. Looks like an extraordinary conference, and I'm very much looking forward to seeing the impact that that has.

So one of the other things I'm very interested in is this intersection of agents—AI agents, multi-agents—and collective intelligence. And as I often say, and you very much manifested in your work, this is not about multi-agent as a stack of different AI agents around. It's saying, well, there are human agents, there are AI agents—so how can you pull these together to get a collective intelligence that manifests the best of both? A group of people and AI working together.

So I'd love to hear about your directions and research in that space.

Pat: Yeah, there is a lot of work that we are doing. And in fact, my PhD advisor, Professor Pattie Maes, is credited as one of the pioneers of software agents. And she is actually receiving the Lifetime Achievement Award in ACM SIGCHI, which is the special interest group in human-computer interaction—this is in a couple of months, actually.

So it's awesome and amazing that she's being recognized as the pioneer of this field.

But the question of agents, I think, is really interesting, because right now, the terminology is very broad. AI is a broad term. AGI is an even broader term. And “agent”—I don't know what the definition is, right?

I mean, some people argue that it's a type of system that can take action on behalf of the user, so the user doesn't need to supervise. This means doing things autonomously. But there are different degrees of autonomy—like things that may require human approval, or things that can just do things on their own. And it can be in the physical world, or the digital world, or in between, right?

So the definition of agent is pretty broad. But I think, again, going back to the question of what is the human experience of interacting with this agent—are we losing our agency or the sense of ownership?

We have many projects that look into and investigate that.

For example, in one project, we design new form factors or new interaction paradigms for interacting with agents. This is a project we worked on with KBTG, which is one of the largest banks in Asia, where we're trying to help people with financial decisions.

If you ask a chatbot, you need to pass back and forth a lot of information—like you need a bank statement, or your savings, or all these accounts. A chatbot is not the right modality.

You could have an AI agent that interacts with people in the task—like if you're planning your financial spending, or investment, or whatever. The AI could be another hand or another pointer on screen. You have your pointer, right? But the AI can be another pointer, and then you can talk to that pointer, and you can feel like there are two agents interacting with one another.

And we showed that—even just changing, using the same exact model—but changing the way that information is flowing and visualized to the user, and the way the user can interact with the agent, rather than going from one screen, then going to the chatbot, typing something, and then going back…

Now, the agent has access to what the user is doing in real time. And because it's another pointer, it can point and highlight things that are important at the moment to help steer the user toward things that are critical, or things they should pay attention to, right?

We showed that this type of interaction reduces cognitive load and makes people actually enjoy the process even more.

So I think the idea of an agent is not a system by itself. It's also the interaction between human and agent—and how can we design it so that it feels like a collaborative, positive collaboration, rather than a delegation that feels like people are losing some agency and autonomy, right?

So I think this is a really, really important question that we need to investigate. Yeah?

Ross: Well, the thing is, it is a trust—a relationship of trust, essentially. So you and it. So there's the nature of the interface between the human, who is essentially trusting an agent—an agent to act on their behalf—and they're able to do things well, that they're able to represent them well, that they check nothing’s missed.

And so this requires a rich—essentially, in a way—emotional interface between the two. I think that's a key part of that when we move into multi-agent systems, where you have multiple agents, each with their defined roles or capabilities, interacting.

This comes, of course—MIT also has a Center for Collective Intelligence. I mean, I'd love to sort of wonder what the intersections between your work and the Center for Collective Intelligence might be.

Pat: Well, one thing that I think both of our research groups focus on is the idea of intelligence not as things that already happen in technologies, but things that happen collectively—at the societal level, or at the collective level.

I think that should be the ultimate goal of whatever we do, right? You should not just make the machine more intelligent, but how do we enhance the overall intelligence?

And I think the question also is: how do we diversify human intelligence as well, right? Because you can be intelligent in a narrow area, but in the real world, problems are very complex. You don't want everyone to think in the same way.

I mean, there are studies showing that on the individual level, AI can make people’s essays better. But if you look across different essays written by people assisted by AI, they start to look the same—which means that there is an individual gain, but a collective loss, right?

And I think that’s a big problem, right? Because now everyone is thinking in the same way. Well, maybe everyone is a little bit better, but if they're all the same, then we have no diverse solution to the bigger problems.

So in one project that we looked into is how do we use AI that has the opposite value as a person—to help make people think more diversely.

If you like something, the AI could like the other thing, and then make the idea something in between. Or, if you are so deep into one thing, the AI could represent the broader type of intelligence that gets you out of your depth, basically.

Or, if you are very broad, maybe the AI will go in deep in one direction—so complementing your intelligence in a way.

And we have shown that this type of AI system can really drive collaboration in a direction that is very diverse—very different from the user.

But at the same time, if you have an AI that is similar to the person—like has the same value, same type of intelligence—it can make them go even deeper. In the sense that if you have a bias toward a certain topic, and the AI also has a bias in the same topic as you, it can make that go even further.

So again, it’s really about the interaction—and what type of intelligence do we want our people to interact with? And what are the outcomes that we care about, whether it's individual or collective?

I think these are design choices that need to be studied and evaluated empirically. Yeah.

Ross: That's fantastic. I mean, I have a very deep belief in human uniqueness. I think we're all far more unique than almost anybody realizes. And society basically makes us look and makes us more the same.

So AI is perhaps a far stronger force in sort of pulling us together—society already is that, yeah. But I mean, to that point of saying, well, I may have a unique way of thinking, or just unique perspectives—and so, I mean, you're talking about things where we can actually draw out and amplify and augment what it is that is most unique and individual about each of us.

Pat: Right, totally. And I mean, I think the former CEO of Google, right, he has said at one point that, why would an individual—why would a person—want to talk to another person when you can talk to an AI that is 100,000 million people at the same time, right?

But I feel like that’s a boring thing. Because the AI could take on any direction. It doesn't have an opinion of its own, right?

But because a human is limited to our own life experience until that point, it gives us a unique perspective, right? When things are everything, everywhere, all at once, it's like generic and has no perspective of its own.

I think each individual person—whether it's the things they’re living through, things that influence their life, things they grew up with—has that sort of story that made them unique. I think that's more— to me, that is more interesting, and I think it's what we should preserve, not try to make everything average out.

So for me, this is the thing we should amplify.

And again, I talk a lot about human-AI interaction, because I feel like the interaction is the key—not just the model capability, but how it interacts with people. What features, what modality it actually uses to communicate with people.

And I think this question of interaction is so interdisciplinary. You need to learn a lot about human behavior, psychology, AI engineering, system design, and all of that, right?

So I think that’s the most exciting field to be.

Ross: Yeah, It's fantastic. So in the years to come, what do you find most exciting about what the Augmenting Humans with AI group could do?

Pat: Well, I mean, many big ideas or aha moments that we want to create—definitely. We have actually an exciting project announcing tomorrow with one of the largest AI organizations or companies in the world. So please watch out for that. There’s new, exciting research in that direction, happening at scale. So there's a big project that's launching tomorrow, which is March 21. So if this is after that, yeah.

I think one thing that we are working on is—we're collaborating with many organizations, trying to focus and make them not just think about AGI, but think about HGI: Human General Intelligence. You know, what would happen to human general intelligence? We want everyone to flourish—not machines to flourish. We want people to flourish, right? To kind of steer many of the organizations, many of the AI companies, into thinking this way.

And in order to do that, we first need a new type of benchmark, right? We have a lot of benchmarks on AI capabilities, but we don't have any benchmarks on what happens to people after using the AI, right? So we need new benchmarks that can really show if the AI makes people depressed, empowers, or enhances these human qualities—these human experiences. We need to design new ways to measure that, especially when they’re using the AI.

Second, we need to create an observatory that allows us to observe how people are evolving—or co-evolving—with AI around the world. Because AI affects different groups of people differently, right? We had a study showing that—this is kind of funny—but people talk about AI bias, that it’s biased toward certain genders, ethnicities, and so on. We did a study showing that, if you remove all the factors, just by the name of people, the AI will have a bias based on the name—or just the last name, right? If you have a famous last name, like Trump or Musk, the AI tends to favor those people more than people who have a generic or regular last name. And this is kind of crazy to me, because you can get rid of all the demographic information that we say causes bias, and just the name of a person already can lead to that bias.

So we know that AI affects people differently. We need to design this type of observatory that we will deploy around the world to measure the impact of AI on people over time—and whether that leads to human flourishing or makes things worse. We don't have empirical evidence for that right now. People are in two camps: the optimistic camp, saying AI is going to bring prosperity, we don’t need to care, we don’t need to regulate. And another group saying AI is going to be the worst thing—existential crisis, human extinction. We need to regulate and kill and stop. But we don't have real scientific empirical evidence on humans at scale.

So that's another thing that MIT’s Advancing Human-AI Interaction is going to do. We're going to try to establish this observatory so that we can inform people with scientific evidence.

And finally, what I think is the most exciting thing: right now, we have so many papers published on AI—more than any human can read, maybe more than any AI can be trained on. Because every minute there's a new paper being published, right? And people are not knowing what is going on. Maybe they know a little bit about their area, or maybe some papers become very famous, but we want to design an Atlas of Human-AI Interaction—a new type of AI for science that allows us to piece together different research papers that come out so that we have a comprehensive view of what is being researched.

What are we over-researching right now? We had a preliminary version of this Atlas, and we showed that people right now do a lot of research on trust and explanation—but less so on other aspects, like loneliness. For example, that AI chatbots might make people lonely—very little research has gone into that.

So we have this engine that's always running. When new papers are being published, the knowledge is put into this knowledge tree. So we see what areas are growing, what areas are not growing, every day. And we see this evolve as the research field evolves. Then I think we will be able to have a better comprehension of when AI leads to human flourishing—or when it doesn't—and see what is being researched, what is being developed, in real time.

So these are the three moonshot ideas that we care about right now at MIT Media Lab. Yeah.

Ross Dawson: Fantastic. I love your work—both you and all of your colleagues. This is so important. I'm very grateful for what you're doing, and thanks so much for sharing your work on The Amplifying Cognition Show.

Pat Pataranutaporn: Thank you so much. And I'm glad that you are doing this show to help people think more about this idea of amplifying human cognition. I think that's an important question and an important challenge for this century and the future century as well.

So thank you for having me. Bye.

The post Pat Pataranutaporn on human flourishing with AI, augmenting reasoning, enhancing motivation, and benchmarking human-AI interaction (AC Ep82) appeared first on Humans + AI.

"We wanted to see what the effect of AI might be on forecasting accuracy… to our surprise, we find that even when the model gives biased or noisy advice, human forecasters still improve—something we didn’t expect."

– Philipp Schoenegger

"I kind of call these Gen AI systems a mirror. Pose it a question, play with scenarios, and see what comes out. It’s like an accelerant for thinking—pushing the boundaries of what’s possible."

– Nikolas Badminton

"Future thinking is an everyday practice. It’s about becoming more aware of what’s happening around us, sensing signals, and collectively imagining what’s next."

– Sylvia Gallusser

"The question of the future isn’t ‘How creative are you?’ but ‘How are you creative?’ Because what we can imagine, we can create—and we have a responsibility to build a better future."

– Jack Uldrich

Philipp Schoenegger is a researcher at London School of Economics working at the intersection of judgement, decision-making, and applied artificial intelligence. He is also a professional forecaster, working as a forecasting consultant for the Swift Centre as well as a 'Pro Forecaster' for Metaculus, providing probabilistic forecasts and detailed rationales for a variety of major organizations.

Nikolas Badminton is the Chief Futurist of the Futurist Think Tank. He is a world-renowned futurist speaker, award-winning author, and executive advisor, with clients including Disney, Google, J.P. Morgan, Microsoft, NASA, and many other leading companies. He is author of Facing Our Futures and host of the Exponential Minds podcast.

Sylvia Gallusser is Founder and CEO of Silicon Humanism, a futures thinking and strategic foresight consultancy. Previous roles include a variety of strategic roles at Accenture, Head of Technology at Business France North America, General Manager at French Tech Hub, and Co-founder at big bang factory. She is also a frequent keynote speaker and author of speculative fiction.

Jack Uldrich is a leading futurist, author, and speaker who helps organizations gain the critical foresight they need to create a successful future. His work is based on the principles of unlearning as a strategy to survive and thrive in an era of unparalleled change. He is the author of 9 books including Business As Unusual.

Websites:

Nikolas Badminton

Nikolas Badminton

Sylvia Gallusser

Jack Uldrich

University Profile:

Philipp Schoenegger

LinkedIn Profile:

Philipp Schoenegger

Nikolas Badminton

Sylvia Gallusser

Jack Uldrich

People

Books & Publications

Technical Terms & Concepts

Ross Dawson: Now, it's wonderful to see the work which you're doing. Speaking of which, recently, you were the lead author of a paper, AI-Augmented Predictions: LLM Assistants Improve Human Forecasting Accuracy.

So first of all, perhaps just describe the paper at a high level, and then we can dig into some of the specifics.

Philipp Schoenegger: Yeah. So the basic idea of this paper is: how can we improve human forecasting?

Human judgmental forecasting is basically the idea that you can query a bunch of very interested and sometimes laypeople about future events and then aggregate their predictions to arrive at surprisingly accurate estimations of future outcomes.

This goes back to work on Superforecasting by Philip Tetlock, and there are a lot of different approaches on how one might go about improving human prediction capabilities.

There might be some training—it was called The Ten Commandments of Forecasting—on how you can be a better forecaster. Or there might be some conversations where different forecasters talk to each other and exchange their views.

And we want to look at how we can—how we could—think about improving human forecasting with AI.

I think one of the main strengths of the current generation of large language models is the interactive nature of the back and forth, having a highly competent model that people can interact with and query whenever they want really.

They might ask the model, "Please help me on this question. What's the answer?" They might also just say, "Here's what I think. Please critique it.

And so this opens up for human forecasters a whole host of different interactions, and we wanted to see what the effect of this might be on forecasting accuracy.

Ross: So that's fascinating. I suppose one of the starting points is thinking about these forecasters. So I suppose, just so people can be clear, human forecasting in complex domains is superior to AI forecasting because they don't have those capabilities.

So now you're saying humans are better than AI alone, but now the results of the paper suggest that humans augmented by AI are superior to either humans alone or AI alone.

Philipp: At the current ammount of papers that I have published, yes, but depending on when this airs, there might be another paper coming out that adds another twist to this.

But yes, in early work, we find that just a simple GPT-4 forecaster underperforms a human crowd, and on top of that, it underperforms just seeing 50% of every question.

But in this paper, we find that if we give people the opportunity to interact with a large language model, which in this case was GPT-4 Turbo, and we prompted it specifically to provide super forecasting.

So our main treatment had a prompt that explained The Ten Commandments of Superforecasting and instructed the model to provide estimates that take care of the base rate.

So you look at how often things like this have typically happened, quantify uncertainty, and identify branch points in reasoning.

But then we also looked at what happens if the large language model doesn't give good advice. What if it gives what we call biased advice? It might be more noisy advice.

So what if the model is told to not think about the base rate—not think about how often things like this happen—to be overconfident, to basically give very high or very low estimates, and be very confident?

And to our surprise, we find that actually, these two approaches similarly effectively improve forecasting accuracy, which is not what we expected.

Ross: So I think that this is a really interesting point because, essentially, this is about human cognition.

It is human cognition taking very complex domains and coming up with a forecast of a probability of an event or a specific outcome in a defined timeframe.

So in this case, the interaction with the AI is a way of enhancing human cognition—they are basically making better sense of the world.

And I guess one of the things that is more distinctive about your approach is, as you say, you could allow them to use anything, any ways of interacting, as opposed to a specific dynamic.

So in this case, it was all human-directed. There was no AI direction. It is AI as a tool, with humans, I suppose, seeking to augment their own ways of thinking about this challenge.

Philipp: Yes, that's right.

And, of course, being human, the vast majority—at least a sizable amount—of participants simply asked the model a question, right?

They just said, "Well, what's the question? What would be the closing value for the Dow Jones at the end of December?" and they just copied it in and saw what the model did.

But then many others did not, and they had their own view. They typed in, "Well, I think that's the answer. What do you think?" or "Please critique this."

And I think these kinds of interactions are especially promising going forward because there's also this whole literature on the different impact of AI augmentation on differently skilled participants, differently skilled workers.

In my understanding, the literature is currently mixed, with studies finding different results.

We didn't find a specific effect here, but other work finds that when the model just gives the answer, low performers typically tend to do better because, you know, they take a lot from the answer, and the model is probably better than them.

But if the model is instructed to give guidance only, low performers tend to not be able to pick up on the guidance and follow it.

But I think there is still a lot of interesting work to be done before we can pin this down because there's so much diversity in which models are being used.

Nikolas Badminton: I do a lot of research on, with every key now, I into a ton of clients. You know, on the client side, I go into the industry. I call people in the industry. I read a ton of academic research behind the industry—stuff on the edge academically, as well as sort of what's in the mainstream and what's being done.

And also, you know, those sort of edge players. When I start to move forward and start to create some new thoughts, then I can sort of start to play around with scenarios. And this is what's become really interesting to me.

I know that you talk a lot about the augmentation of capability through the use of things like generative AI and the such like. This has been something that I've been playing with quite a lot—not only from the generation of textual content but also the exploration from a visual perspective as a helping mechanism to take us in whole new directions as well.

I mean, in my work, it's like signals to trends, to scenarios, and to stories. I've really been trying to push the boundaries of what scenario exploration is with platforms like ChatGPT, Claude, and Gemini, and starting to see what we can do to look at positive and dystopian scenarios, which was obviously part of the work that I was doing, a part in Facing Our Futures.

Over the last couple of years, since that book was completed, zero Gen AI sort of help, as it was in my book. And actually, very little Gen AI help is going to be in my next book because, contractually, you're not allowed to do this.

So what we have—what we can do—is start to explore the mirror. I kind of call these Gen AI systems a mirror. Pose it a question. Pose it some scenarios. Try to work out and see what comes out of it.

And generally, what I find is maybe I'm talking about energy and ecological ecosystems, and I'll pose a question, "What if renewable energy is pushed to the side, green initiatives are canceled, and we go full tilt into a maximalist fossil fuel society?"

In preparation for this chat, I went into that to delve even deeper into the mechanisms behind that. And it's sort of interesting—you get this mirror of like, "Oh yeah, I kind of expect that, you know, the answers to come from that."

Okay, let's push that out to 2050. Yeah, it's kind of an accelerant and whatever. It's kind of interesting when you start to think about the reference points of all these systems and where they're getting it from.

Where something like Claude and ChatGPT actually feels like they've been drinking from the same fountain, and Gemini just seems to be a little bit freaky.

So it's super interesting. As I went into it, it was like poetic and dystopic.

For example, I asked this: "Describe a world in 2100 where environmentally friendly, non-carbon fuel solutions are discarded."

And I went on and on in a prompt, very directional. The others would be like, "Here's a list of things that happen"—very cold. I didn't ask it to write in a particular style of a publication or anything like that.

And then Gemini just came out with this. And this is fabulous:

"The year is 2100. The gamble on renewables failed spectacularly. Big Oil, whispering sweet nothings of energy independence and economic growth, won the hearts and minds of a desperate world. The result? A planet drowning in its own fumes."

And I kind of love that poetic nature.

Gemini, I think, is sort of the unsung hero a little bit, right? In the scheme of things, suddenly, we're getting something interesting that starts to talk about the geopolitical chessboard, tech on steroids, violence, and exodus.

And it's like—whoa.

Ross: A lot of it, I think, is about sensitizing ourselves to signals so that we are more likely to notice the things that are relevant or important or point to things that might change in the future.

And that's what futurists do. But how can we, I suppose, convey this as a capability or skill that others can learn and develop—that’ll been able to see and sense signals that, you know, point to change?

Sylvia Gallusser: It's a very interesting thing with signals. It's like raw material. It's something that anybody can apprehend, and that's what makes future thinking something that really anybody can work with and develop as a personal skill.

Because it's about becoming more aware of what is going on around us. And that's why I think it works really in tandem, in deal with the first step, which is about knowing always more, understanding always more about what is the long-term landscaping, and then being more aware of the variation.

And this can go from analyzing behaviors of people around you—like, what changed during the pandemic? Were people more polite, more civilized? Did we see new behaviors, new words?

Maybe also studying popular culture is a very interesting aspect because if you see what is going on in the media—TV series, movies, books—you also sense a lot of what people are attracted to. What new changes are starting when there's this kind of enthusiasm for a new book; sometimes, that means something.

So how can you get more aware of this? It's really an everyday practice, and I like to say two things: it's a personal practice, and it's a collective practice.

That's something you can really train yourself to do all the time—just reading the news, being aware of what is around you, just having your sensors open to the world around. And once again, it's all senses. It's about listening. It's about observing people around you. It's a different taste in the air. It's really multi-sensitive here.

Why I say it's also collective is that, you know, the futurist community is very active. It's not that big; it's small. But it's very interconnected.

And there are a lot of platforms to be able to exchange around signals. They call it sometimes signal swarming or signal scanning—you have different names for it—but the idea is that futurists love to exchange around that topic, to meet and say, "Hey, this week, what did you notice?"

And once again, this STEEPLE aspect is interesting because when you're on your own, coming maybe from one industry or one profession, maybe you're a kind of a bias around one or the other.

Like, I'm coming from technology, so at first, I would really focus on everything around new technology and so on. But I guess someone who's a psychologist might have a different opinion. An economist might see things differently.

So coming together as a collective, as a community, is really interesting into enhancing and amplifying the way you connect with those signals around you.

And finally, I would say, on top of it being collective, what's interesting when you want to bring a group, a population, a company, or a corporation to work around future thinking is to build the capability to do this.

It's very simple. It can start with just an Excel file. It doesn't need something very fancy.

But just bring people to come to see what signals are and get them to understand the texture of it—how does it look like? How does it sound like? And they start to log on their own signals.

And then you already have a big bases of signals of change in a corporation. A great first way to enter the field of foresight.

Ross: So one of the other things you were talking about was putting yourself in the scenario.

And I suppose part of the practice is to create a useful scenario that thus helps you think about new things or envisage things that help shape your current actions.

But as individuals, what are ways in which we can, I suppose, conceive of and bring ourselves—or enter into—I think you used the word meditation there.

And, you know, I’d love to hear about that. What is that practice? How do we put ourselves, immerse ourselves in these useful future scenarios?

Sylvia: Absolutely. Once again, you know, it can be very personal and intimate, or it can be something more collective.

So I try to address both aspects because I think they can work really well together. You can develop your own future-thinking practice as an everyday discipline, let's say.

I wrote a few years ago, an article about mental stretching exercises you can practice to work on that. It can go from dealing with different perspectives, trying to develop empathy, putting yourself in the shoes of someone else, and imagining a story.

You know what? Actually, learning new languages and learning new cultures is also a great way to practice this perspective change and teasing things in different ways.

Reading, listening, and learning about fiction, for me, has been an immense way to stretch myself to see futures that are possible and not necessarily dystopian.

That's why I love to talk about science fiction, because we tend to think, to see science fiction as something very dystopian and very scary and not necessarily the good way to start for people who are scared about the future.

But I would say there are more and more interesting science fiction now that create a future world that is not necessarily negative. They can be really engaging and develop a plot which has a narration where the problems are, but it doesn't mean that the negative aspect is the world-building.

Like the story, to be interesting, needs to always have something of a dilemma or something of a complexity or a knot to it.

But it can be interpersonal stories, not necessarily in the world-building around it.

So I think science fiction and future fiction really offer us ways to think about the future.

So, for example, the way we do it collectively with groups, and I was talking about those meditative exercises.

A really great way we've been doing it in the past was around the future of the home.

Because during the pandemic, the home evolved dramatically, and not just the structure but also the way we reorganized life within it.

And I like to talk about the structures and the intangibles that happen in the home.

So what we would do, for example, in terms of envisioning meditations with a few groups, was really you waking up in the future home you live in—maybe 10 years from now, 20 years from now.

How do you wake up? What is the first trigger? What happens?

Is it a wake-up call? Is it natural lighting? Do you still live in a bedroom?

Like, we really start just—what do you smell? What do you think? What do you feel? How does it sound?

So five senses meditation is really effective.

Changing perspective, as I was saying, and so on.

So these are different tools we would use to bring people to get into that state of the future and then go throughout a day in the life.

Like, okay, what do you do from your bed? Then do you go to breakfast? Do you go to your bathroom?

How does the bathroom look? Is it interactive? Do you live alone? Do you live with other people in a community?

And just—it starts asking so many questions that people naturally get their minds to wander around the future home.

And that was a really great tool to get a sense of that new type of space that could exist.

And, oh, they would like that home to be.

Because, once again, it is also about developing what would be our preferable future, our favorite futures, and building them.

Jack Uldrich: And I've spent a lot of time as a futurist with the concept of unlearning.

It's that people in organizations—it’s not that they can't understand the future is going to change. What we have a really difficult time doing is letting go of the way we've always done things.

And so I think when we're talking about the future of work, to me, work does give most humans this intrinsic value, and they feel as though they're an integral part of a community.

And so I think there will always be this innate need to be doing something—not just for yourself but on behalf of something bigger.

And when I say bigger, typically I'm thinking of community. You just want to do something for, of course, yourself, your immediate family, but then your neighborhood and your community.

And so as I think about the long-term future, one of the things I'm really excited about is—first, I'm going to go dark, but I think there's going to be a bright side to this.

One of the things that I think is happening right now that's not getting enough attention, as a futurist, is that the internet is breaking.

In the sense that there's so much misinformation and disinformation out there that we can no longer trust our eyes and our ears in this world of artificial intelligence.

And I think that's going to become increasingly murkier, and it's going to be really destabilizing to a lot of people and organizations.

So what's the one thing we still can trust? What's small groups that are right in front of us?

And so I think one of the things we're going to see in a future of AI is an increased importance on small communities.

There's some really compelling science that says the most cohesive units are about 150 people in size.

And this is true in the military, educational units, and other things like that.

And I think that we might start seeing that, but it's going to look different than in the past.

Like, I'm not suggesting that we're all going to look like Amish communities here in the U.S., where we're saying no to technology and doing things the old-fashioned way.

But the new communities of the future are—and now I'm just thinking out loud—something I want to spend more time thinking about.

Like, what will that look like? What will the roles and the skills be needed in this new future?

And again, I don't have any answers right now, just more questions and thinking.

But it's one of these scenarios I could see playing out that might catch a lot of people by surprise.

Ross: Yeah, very much so. I mean, we are a community-based species, and the nature of community has changed from what it was.

And I think, you know, thinking about the future of humanity, I think a future of community and how that evolves is actually a very useful frame to round out.

Jack, what advice can you share with our listeners on how to think about the future? I suppose you did a little at the beginning.

But, I mean, do you have any concluding thoughts on how people can usefully think about the extraordinary change in the world today?

Jack: Yeah, the first thing I would say is this—and I was just doing a short video on this.

Ever since we've been in grade school, most of us have been asked the question or graded on the question of How creative are you?

And if you ask most people, like on a scale of one to ten, to just answer that question, they'll do it.

But you know what I always tell people? That's a bad question.

The question of the future isn't How creative are you? It is How are you creative?

Each and every one of us is creative in our own way. And as a futurist, I take that really seriously.

We do have the ability to create our own future, but we first have to understand that we are creative, and most people don't think of themselves that way.

So how do you nurture creativity?

And this is where I'm trying to spend a lot of my time as a futurist. This is where the ideas of unlearning and humility come in.

But I would say it starts with curiosity and questions, and that's why I like getting out under the night stars and just being reminded of how little I actually know.

But then, it's in that space of curiosity that imagination begins to flow.

And there's this wonderful quote from Einstein—most people would say he was one of the more brilliant minds of the 20th century. He said, Imagination is more important than knowledge.

Like, why did Einstein, this great scientist, say that?

And I think—and I don’t have proof of this—that everything around us today was first imagined into existence.

It was imagined into existence by the human mind.

The very first tool. The very first farm implement.

And then farming as an industry, and then civilizations and cities and commerce and democracy and communism.

They were all imagined first into existence.

And so, what we can imagine, we can, in fact, create.

And that's why I'm still optimistic as a futurist—this idea that we're not passive agents, that we can create a future.

And I just like to remind people that our future can, in fact, be incredibly fucking bright.

The idea that we can have cleaner water and sustainable energy and affordable housing and better education and preventive health care.

We can address inequality. We can address these issues.

People just have to be reminded of this.

And so, at the end of the day, that's why I get fired up, and I don’t think I’ll ever sort of lose the title of futurist, because until my last breath, I’m going to be, hopefully, reminding people that we can create—and we have a responsibility to create—a better future.

Let me just end on this.

I think the best question we can ask ourselves right now comes from Jonas Salk, the inventor of the polio vaccine.

And he said, Are we being good ancestors?

And I think the answer right now is, we're not.

But we still have the ability to be better ancestors.

And maybe if I could just say one last thing—I also spend a lot of time helping people just embrace ambiguity and paradox.

And here’s the truth: the world is getting worse.

In terms of climate change, the rise of authoritarianism, inequality—you could say things are going bad.

But at the same time, on the other hand, you could say the world is getting demonstrably better.

It has never been a better time to be alive as a human.

The likelihood that you’re going to die of starvation or war or not be able to read—never been lower.

So the world is also getting better.

But the operative question becomes: How can we make the world even better?

And that’s where we have to spend our time.

And that’s why we need creativity, curiosity, and imagination—to create that better future.

The post Amplifying Foresight Compilation (AC Ep81) appeared first on Humans + AI.

"AI can make the process of sensing for signals much faster and much more efficient. You can think of it as a supplement to our brain. It can sort through massive amounts of data, track the latest developments, and flash alerts when something important emerges."

– Rita McGrath

"What I found surprising in our exercises was how disruptive AI was. At first, I thought they would hate it, but they actually liked it. It made them stop and think because it forced them to break out of their usual patterns and consider ideas they wouldn’t have consciously introduced into the discussion."

– Christian Stadler

"AI can accelerate the foresight process. It can help generate diverse perspectives, identify second-degree impacts, and uncover biases we might not notice. Of course, human critical thinking is still essential—we shouldn’t accept AI outputs as absolute truth, but rather use them as a starting point."

– Valentina Contini

"One key area where AI excels is handling cognitive complexity. Humans struggle to hold thousands of variables in their heads, but AI can process vast amounts of interconnected data. The challenge is designing interfaces that allow humans to interact with this complexity in an intuitive way."

– Anthea Roberts

Rita McGrath is one of the world’s top experts on strategy and innovation. She is consistently ranked among the top 10 management thinkers globally and has earned the #1 award for strategy by Thinkers 50. She is Professor of Strategy at Columbia Business School, and Founder of the Rita McGrath Group and Valize LLC. Her books include The End of Competitive Advantage and Seeing Around Corners.

Christian Stadler is a professor of strategic management at Warwick Business School. He is author of Open Strategy, which was named as a Best Business Book by Financial Times and Strategy + Business and has been translated into 11 languages. His work has been featured in Harvard Business Review, New York Times, Wall Street Journal, CNN, BBC, and Al Jazeera, among others.

Valentina Contini is an innovation strategist for a global IT services firm, a technofuturist, and speaker. She has a background in engineering, innovation design, AI-powered foresight, and biohacking. Her previous work includes founding the Innovation Lab at Porsche.

Anthea Roberts is Professor at the School of Regulation and Global Governance at the Australian National University (ANU) and a Visiting Professor at Harvard Law School. She is also the Founder, Director and CEO of Dragonfly Thinking. Her latest book, Six Faces of Globalization, was selected as one of the Best Books of 2021 by The Financial Times and Fortune Magazine. She has won numerous presitigious awards and has been named “the world's leading international law scholar” by the League of Scholars.

Websites:

Rita McGrath

Rita McGrath

Christian Stadler

Valentina Contini

Anthea Roberts

Anthea Roberts

University Profile:

Rita McGrath

Christian Stadler

Anthea Roberts

LinkedIn Profile:

Rita McGrath

Christian Stadler

Valentina Contini

Anthea Robert

Companies & Organizations

Technical Terms & AI-Related

Ross Dawson: One of the key themes is strategy. How do we do strategy in a world that is accelerating, with all these overlay themes? There are, as you say, 10x shifts in many dimensions of work. This brings us to human capabilities. Humans have limited, finite cognition, even though we have extraordinary capabilities far transcending anything else. Now, we have AI to augment, support, or complement us.

I’d like to dive in deep, but just to start—what is your framing around human capabilities in strategic thinking today, and how they are complemented by AI?

Rita McGrath: Sure. Well, as I mentioned, human brains think in linear terms. We think immediately in terms of getting from here to there to avoid a predator. Back in the day when we were evolving, that worked pretty well. But we don’t do very well with exponential systems because they look small, and they look small, and they go small—until suddenly they don’t. It’s the whole “gradually, then suddenly” idea.

What I argue is that you need to supplement what your brain can manage on its own. This is where I think AI comes in. What I’ve set up with companies is a series of what I call “time zero events,” which signal that a future inflection point has arrived. We don’t know exactly when, but we work backward and ask, “Before that happens, what would have to be the preceding situations?”

AI can make that process of sensing for signals much faster and much more efficient. You can think of it as a supplement to our brain. It can sort through massive amounts of data, track the latest developments, and flash alerts when something important emerges. This allows us to blend human imagination—something AI is not very good at—with AI’s ability to crunch massive amounts of data. That’s where I think AI will have a lot of power in strategy.

Ross: One of the core themes of my work, and I think yours as well, is sense-making. We have vast amounts of information out there. As strategists, we need to take in that information, make sense of it, and make effective decisions as a result. How can AI support our ability to comprehend how the world is working so that we can make better decisions?

Rita: AI is really good at taking large amounts of information and breaking it into digestible chunks. Humanity has limits to how much information it can process. There’s actually a whole line of theory on this, which states that search, in the traditional sense, is not costless. Theoretically, a rational human being would entertain every possible combination of possibilities, create decision criteria, and then select the best option. But humans have cognitive limits, whereas machines have far fewer.

Properly instructed, AI can present us with different pictures of the world. Another thing humans aren’t very good at is generating variety. Think of those old creativity exercises where someone asks you to come up with as many uses as possible for a paperclip. People start with obvious answers: “It can hold papers together,” “It can mark your place in a book,” “It can unlock things.” But after 50 or 60 uses, they run out of steam. Many ideas are anchored on the first few.

Machines, on the other hand, don’t have those biases. They might generate 300 possible uses—sure, 200 of them might be terrible ideas, but they would be more divergent than what humans come up with. That’s where AI helps in sense-making. It shows us possibilities we wouldn’t have seen otherwise.

Ross: Now, let’s dig into how AI can be used in the boardroom. One way that resonates with board directors is “red teaming,” where you have a decision and ask AI to generate counterarguments. AI can surface concerns that might not come up in human discussions. What other applications have you found valuable for AI in the boardroom?

Christian Stadler: What I found surprising in our exercises was how disruptive AI was. Imagine a group of people who have worked together for a long time. Their discussions are smooth because they know how each other thinks. Then, I introduce ChatGPT into the meeting.

I’d tell them, “Read these five pages,” and suddenly, they’re confronted with a long list of new insights. It disrupted their usual flow. At first, I thought they would hate it, but they actually liked it. It made them stop and think. The disruption forced them to break out of their usual patterns and consider ideas they wouldn’t have consciously introduced into the discussion.

Ross: What are the ways in which you are seeing or applying tools to augment the foresight process?

Valentina Contini: I started looking into this about two years ago, when GPT-3.5 was released. One of the things that frustrated me was that generating scenarios for companies took too long. You needed to involve multiple experts and stakeholders, which meant it only happened every three to five years. But in today’s rapidly changing world, that’s not enough.

AI can accelerate the foresight process. It can help generate diverse perspectives, identify second-degree impacts, and uncover biases we might not notice. It’s especially useful in tools like a futures wheel, where many perspectives need to be mapped. AI can bring in unexpected viewpoints based on large-scale data analysis. Of course, human critical thinking is still essential—we shouldn’t accept AI outputs as absolute truth, but rather use them as a starting point.

Ross: Human-AI collaboration involves complex problems where humans retain the highest-level context and decision-making ability, while AI complements our cognition. What does that interface look like?

Anthea Roberts: This is one of the most fascinating questions of our time. Both humans and AI have different strengths, and the way we interact with AI is evolving.

For example, when working with large language models, humans shift from being primary generators of content to being managers and editors. We direct how the AI works and refine its outputs. This requires metacognition—not just thinking about our own thinking, but also understanding how the AI thinks.

One key area where AI excels is handling cognitive complexity. Humans struggle to hold thousands of variables in their heads, but AI can process vast amounts of interconnected data. The challenge is designing interfaces that allow humans to interact with this complexity in an intuitive way. A simple chat interface isn’t enough—we need tools that allow for narrowing focus, cognitive offloading, and iterative collaboration.

Another challenge is balancing AI’s overwhelming amount of information with human discernment. Many people feel deluged by AI-generated content, making it crucial to develop skills for filtering and applying insights effectively.

Ross: So AI not only provides information but also changes the way we think and interact with complexity?

Anthea: Exactly. Over the last year and a half, I’ve realized that much of my work is metacognitive. I don’t tell people what to think, but I help them understand how they think. The same applies to AI—we need to recognize its biases, workflows, and limitations while leveraging its strengths.

One of the biggest challenges will be developing interdisciplinary AI agents that can collaborate across different fields of expertise. AI will evolve into an indispensable partner in decision-making, but we need to ensure that humans remain in control of the broader context and ethical considerations. How we navigate this balance will define the future of AI-human collaboration.

The post AI for Strategy Compilation (AC Ep80) appeared first on Humans + AI.

“Collective intelligence is the ability of a group to solve a wide range of problems, and it’s something that also seems to be a stable collective ability.”

– Anita Williams Woolley

"When you get a response from a language model, it’s a bit like a response from a crowd of people. It’s shaped by the collective judgments of countless individuals."

– Jason Burton

"Rather than just artificial general intelligence (AGI), I prefer the term augmented collective intelligence (ACI), where we design processes that maximize the synergy between humans and AI."

– Gianni Giacomelli

"We developed Conversational Swarm Intelligence to scale deliberative processes while maintaining the benefits of small group discussions."

– Louis Rosenberg

Anita Williams Woolley is the Associate Dean of Research and Professor of Organizational Behavior at Carnegie Mellon University’s Tepper School of Business. She received her doctorate from Harvard University, with subsequent research including seminal work on collective intelligence in teams, first published in Science. Her current work focuses on collective intelligence in human-computer collaboration, with projects funded by DARPA and the NSF, focusing on how AI enhances synchronous and asynchronous collaboration in distributed teams.

Jason Burton is an assistant professor at Copenhagen Business School and an Alexander von Humboldt Research fellow at the Max Planck Institute for Human Development. His research applies computational methods to studying human behavior in a digital society, including reasoning in online information environments and collective intelligence.

Gianni Giacomelli is the Founder of Supermind.Design and Head of Design Innovation at MIT’s Center for Collective Intelligence. He previously held a range of leadership roles in major organizations, most recently as Chief Innovation Officer at global professional services firm Genpact. He has written extensively for media and in scientific journals and is a frequent conference speaker.

Louis Rosenberg is CEO and Chief Scientist of Unanimous A.I., which amplifies the intelligence of networked human groups. He earned his PhD from Stanford and has been awarded over 300 patents for virtual reality, augmented reality, and artificial intelligence technologies. He has founded a number of successful companies including Unanimous AI, Immersion Corporation, Microscribe, and Outland Research. His new book Our Next Reality on the AI-powered Metaverse is out in March 2024.

Websites:

Gianni Giacomelli

Louis Rosenberg

University Profile:

Anita Williams Woolley

Jason Burton

LinkedIn Profile:

Anita Williams Woolley

Jason Burton

Gianni Giacomelli

Louis Rosenberg

Anita Williams Woolley: Individual intelligence is a concept most people are familiar with. When we're talking about general human intelligence, it refers to a general underlying ability for people to perform across many domains. Empirically, it has been shown that measures of individual intelligence predict a person's performance over time. It is a relatively stable attribute.

For a long time, when we thought about intelligence in teams, we considered it in terms of the total intelligence of the individual members combined—the aggregate intelligence. However, in our work, we challenged that notion by conducting studies that showed some attributes of the collective—the way individuals coordinated their inputs, worked together, and amplified each other's contributions—were not directly predictable from simply knowing the intelligence of the individual members.

Collective intelligence is the ability of a group to solve a wide range of problems. It also appears to be a stable collective ability. Of course, in teams and groups, you can change individual members, and other factors may alter collective intelligence more readily than individual intelligence. However, we have observed that it remains fairly stable over time, enabling greater capability.

In some cases, collective intelligence can be high or low. When a group has high collective intelligence, it is more capable of solving complex problems.

I believe you also asked about artificial intelligence, right? When computer scientists work on ways to endow a machine with intelligence, they essentially provide it with the ability to reason, take in information, perceive things, identify goals and priorities, adapt, and change based on the information it receives. Humans do this quite naturally, so we don't really think about it.

Without artificial intelligence, a machine only does what it is programmed to do and nothing more. It can still perform many tasks that humans cannot, particularly computational ones. However, with artificial intelligence, a computer can make decisions and draw conclusions that even its own programmers may not fully understand the basis of. That is where things get really interesting.

Ross Dawson: We'll probably come back to that. Here at Amplifying Cognition, we focus on understanding the nature of cognition. One fascinating area of your work examines memory, attention, and reasoning as fundamental elements of cognition—not just on an individual level, but as collective memory, collective attention, and collective reasoning.

I’d love to understand: What does this look like? How do collective memory, collective attention, and collective reasoning play into aggregate cognition?

Anita: That’s an important question. Just as we can intervene to improve collective intelligence, we can also intervene to improve collective cognition.

Memory, attention, and reasoning are three essential functions that any intelligent system—whether human, computer, or a human-computer collaboration—needs to perform. When we talk about these in collectives, we are often considering a superset of humans and human-computer collaborations. Research on collective cognition has been running parallel to studies on collective intelligence for a couple of decades.

The longest-standing area of research in this field is on collective memory. A specific construct within this area is transactive memory systems. Some of my colleagues at Carnegie Mellon, including Linda Argote, have conducted significant research in this space. The idea is that a strong collective memory—through a well-constructed transactive memory system—allows a group to manage and use far more information than they could individually.

Over time, individuals within a group may specialize in remembering different information. The group then develops cues to determine who is responsible for retaining which information, reducing redundancy while maximizing collective recall. As the system forms, the total capacity of information the group can manage grows considerably.

Similarly, with transactive attention, we consider the total attentional capacity of a group working on a problem. Coordination is crucial—knowing where each person's focus is, when focus should be synchronized, when attention should be divided across tasks, and how to avoid redundancies or gaps. Effective transactive attention allows groups to adapt as situations change.

Collective reasoning is another fascinating area with a significant body of research. However, much of this research has been conducted in separate academic pockets. Our work aims to integrate these various threads to deepen our understanding of how collective reasoning functions.

At its foundation, collective reasoning involves goal setting. A reasoning system must identify the gap between a desired state and the current state, then conceptualize what needs to be done to close that gap. A major challenge in collective reasoning is establishing a shared understanding of the group’s objectives and priorities.

If members are not aligned on goals, they may decide that their time is better spent elsewhere. Thus, goal-setting and alignment are foundational to collective reasoning, ensuring that members remain engaged and motivated over time.

Ross: One of the interesting insights from your paper is that large language models (LLMs) themselves are an expression of collective intelligence. I don’t think that’s something everyone fully realizes. How does that work? In what way are LLMs a form of collective intelligence?

Jason Burton: Sure. The most obvious way to think about it is that LLMs are machine learning systems trained on massive amounts of text. Companies developing these language models source their text from the internet—scraping the open web, which contains natural language encapsulating the collective knowledge of countless individuals.

Training a machine learning system to predict text based on this vast pool of collective knowledge is essentially a distilled form of crowdsourcing. When you query a language model, you aren’t getting a direct answer from a traditional relational database. Instead, you receive a response that reflects the most common patterns of answers given by people in the past.

Beyond this, language models undergo further refinement through reinforcement learning from human feedback (RLHF). The model presents multiple response options, and humans select the best one. Over time, the system learns human preferences, meaning that every response is shaped by the collective judgments of numerous individuals.

In this way, querying a language model is like consulting a crowd of people who have collectively shaped the model’s responses.

Gianni Giacomelli: I view this through the lens of augmentation—augmenting collective intelligence by designing organizational structures that combine human and machine capabilities in synergy. Instead of thinking of AI as just a tool or humans as just sources of data, we need to look at how to structure processes that allow large groups of people and machines to collaborate effectively.

In 2023, many became engrossed with AI itself, particularly generative AI, which in itself is an exercise in collective intelligence. These systems were trained on human-generated knowledge. But looking at AI in isolation limits our understanding. Rather than just artificial general intelligence (AGI), I prefer the term augmented collective intelligence (ACI), where we design processes that maximize the synergy between humans and AI.

Louis Rosenberg: There are two well-known principles of human behavior: one is collective intelligence—the idea that groups can be smarter than individuals if their input is harnessed effectively. The other is conversational deliberation—where groups generate ideas, debate, surface insights, and solve problems through discussion.

However, scaling these processes is difficult. If you put 500 people in a chat room, it becomes chaotic. Research shows that the ideal conversation size is five to seven people. To address this, we developed Conversational Swarm Intelligence, using AI agents in small human groups to facilitate discussions and relay key insights across overlapping subgroups. This allows us to scale deliberative processes while maintaining the benefits of small group discussions.

The post Collective Intelligence Compilation (AC Ep79) appeared first on Humans + AI.

“I’m cautiously optimistic because never before has technology been as accessible as it is now—being able to interact with machines in a way that feels so natural to us, rather than in ones and zeros or more technical ways. AI shouldn’t replace what exists but augment and enhance our creativity, helping us tap into what makes us uniquely human.”

– Helen Lee Kupp

Helen Lee Kupp is co-founder and CEO of Women Defining AI, a community of female leaders applying and driving AI. She was previously leader of strategy and analytics at Slack and co-founder of its Future Forum. She is co-author of the best-selling book “How the Future Works: Leading Flexible Teams to do the Best Work of Their Lives".

Website:

LinkedIn Profile:

Helen Lee Kupp

People

Companies & Organizations

Technical Terms & Concepts

Ross Dawson: Helen, it is a delight to have you on the show.

Helen Lee Kupp: It's good to be here. I love how we first started talking over an AI research paper. It was very random but awesome.

Ross: Well, that's pushing the edges, trying to find what's out there and see what comes on the other side. AI is emerging, and we’re sitting alongside each other. How are you feeling about today and how humans and AI are coming together?

Helen: I feel cautiously optimistic, and part of that is because I've been in tech for so long. Prior to getting much deeper into AI, I was working on flexible work and research around how to rethink and redesign how we, as humans, collaborate in a way that is more personalized, more customized, and helps more people bring their best selves to work and do their best work.

It was serendipitous that around the same time, there was an increase in AI innovation. Now, we had technology to pair with the equation of redesigning work. COVID forced us to rethink work, not just from a people and process perspective but alongside rapid technological change. I'm cautiously optimistic because never before has technology been as accessible as it is now. We can interact with machines in a way that feels so natural rather than in ones and zeros or technical ways.

Ross: I’m very aligned with that. One of the things you said was "bring your best self to work." I think of it as human potential. If we're creating a future of work, we have potential futures that are not so great and others that are very positive, where people express more of who they are and their capabilities. How can we create organizations like that?

Helen: It starts with recognizing that everyone has different preferences and work styles. Organizations, teams, and leaders need to meet people where they are rather than force them into rigid structures that worked in the past.

I often share this story—I’m deeply introverted. Despite jumping onto this podcast with you, I have always been an introvert. Navigating an extroverted world takes extra energy. In traditional office and meeting environments, I had to work harder to show up. However, when I had more diverse formats to interact with my team and leadership, it unlocked something for me.

Instead of pretending to be the loudest in the room, I could find my own ways of expressing ideas—through text, written formats, or chat. It made work easier for me. When you think about how that manifests across a team, leaders and organizations must avoid putting rigid boxes around collaboration—whether it’s the hours we work or the place where we work. Increasing flexibility enables people to express themselves and bring forward ideas that might otherwise remain hidden.

Ross: That’s a compelling vision. How do you bring that to reality? What do you do inside an organization to foster and enable that?

Helen: One of the tools that helped in our research on the future of work and redesigning organizations is something simple—creating a team operating manual.

The act of explicitly writing down the different ways we interact as a team opens up discussions. It allows for feedback: "Does this work for you? Should we try something different?" When these conversations don’t happen, implied assumptions remain—such as the norm of working in an office from nine to five.

Explicitly stating and questioning these assumptions is step one. Then, organizations should give teams and managers the flexibility to define how they work within their sub-teams. Having operating manuals, sharing what works for your team, and bubbling up insights allow for a more bottom-up approach rather than a top-down one. It treats people like adults who understand their preferences and styles.

Ross: That’s really nice. PepsiCo had an initiative where teams coordinated among themselves to determine their availability and collaboration methods. I wonder if we can push that further. People are often conditioned to fit into roles and adjust to their environments. Can we help people recognize their self-imposed constraints and flourish beyond them?

Helen: This is where I’m cautiously optimistic about AI and how we integrate technology into work. When people start using AI, the initial question is often, "How can I do this more efficiently?" AI is a powerful tool that shortens tasks—like a calculator removing the need for mental math.

However, once people move beyond efficiency, they begin asking, "What can I do differently?" AI allows us to do things we couldn’t before. It helps break conventional thinking. For example, if you use a large language model to generate 10 variations of an idea, it removes emotional bias. It shifts the conversation from defending one perspective to evaluating multiple ideas. This fosters creative discourse and integrates seamlessly into workflows without feeling like extra work.

AI should not replace what exists but augment and enhance our creativity—helping us tap into what makes us uniquely human.

Ross: So, AI helps individuals bring different perspectives and expand their thinking?

Helen: Exactly. One of my favorite things to do with large language models is to open up the funnel. Whether it’s brainstorming writing styles, problem-solving, or scoping solutions, AI presents multiple potential paths. This reminds us that there is no single correct answer—only possibilities to explore.

Ross: Gregory Bateson said wisdom comes from multiple perspectives. We now have multiple perspectives on demand. You work with leaders to redesign organizations. What guidance do you suggest? How can organizations evolve from existing structures?

Helen: I don’t have the perfect answer for what the shape of organizations should be. However, we’ve been transitioning from hierarchical structures to teams-of-teams for a while, with varying success. The biggest challenge is breaking out of our mental paradigms of control.

Flexible work means allowing managers and teams to design their workdays and collaboration methods rather than enforcing a company-wide approach. AI introduces another paradigm shift—it behaves unpredictably compared to traditional technology. Leaders must accept that they don’t have all the answers.

Some of the best AI-driven innovations come from employees who work closely with the technology daily. For example, a data scientist evaluating AI’s role in data processing can quickly identify where it adds value and where it falls short. These innovations emerge at the edges, from individuals experimenting in real time.

Leaders must create environments where experimentation, sharing, and collaboration thrive. Instead of dictating policies top-down, they should spotlight grassroots innovations and scale them across the organization.

Ross: So, you’re describing emergence—where leaders set conditions for innovation rather than dictate precise rules?

Helen: Exactly. Constraints breed creativity. If there are no guardrails or structures, people stick to the status quo and don’t innovate. Leaders must provide the right nudges—whether through hackathons, dedicated experimentation time, or open Slack channels to share discoveries.

Some organizations set up "experiment hours"—weekly meetings where teams explore AI applications in a low-pressure, fun environment. This fosters creativity and keeps innovation moving.

Ross: That’s a great example. Speaking of multiple perspectives, one of your recent ventures is Women Defining AI. What is it about?

Helen: Women Defining AI started as an experiment about a year and a half ago. I had been working with generative AI models and noticed a significant gender gap in AI adoption. Data showed men adopting AI at higher rates than women, and anecdotally, I saw the same trend.

Initially, it was just a study group where I shared what I was learning with other women. Within days, 50 people joined, and by month two, we had 150 members. It became clear that women wanted a space to ask questions, learn together, and experiment without judgment.

Now, Women Defining AI is a virtual community that helps women at different stages of their AI journey. Whether it’s understanding AI’s role in their work, automating tasks, or building solutions, we guide them in gaining technical confidence and shaping the field. Some members have landed AI-related jobs or joined AI policy teams at their organizations.

Having diverse perspectives in AI is crucial. Women in our community, particularly those from HR and other industries, quickly identify biases and blind spots that might otherwise go unnoticed. We need more voices questioning and shaping AI while we’re still in its early stages.

Ross: That’s fantastic. Looking ahead to 2026, what excites you most?

Helen: Personally, I’m excited about having our third baby! It’s a reminder of the new perspectives each generation brings.

For Women Defining AI, 2025 will be the year we build in public. We’ve been experimenting and learning internally, but now we’re sharing real stories and projects to inspire more builders and technologists.

Ross: That’s fantastic. Thank you for your time, insights, energy, and passion.

Helen: Thanks for having me.

The post Helen Lee Kupp on redesigning work, enabling expression, creative constraints, and women defining AI (AC Ep78) appeared first on Humans + AI.

"Generative AI is the first technology with an almost natural propensity to build a symbiotic relationship with us. But symbiosis isn’t always mutualistic—it can be parasitic, where AI benefits at the detriment of humans. How we deploy AI will determine which path we take."

– Alexandra Diening

"AI provides dual affordances—it can automate our work or augment our abilities. The key challenge is deciding where to draw the line. In low-stakes tasks, automation makes sense. But in high-stakes decision-making, human intuition is irreplaceable."

– Mohammad Hossein Jarrahi

"We talk a lot about lifelong learning, but we also need to embrace lifelong forgetting. If we keep piling new knowledge on top of outdated thinking, we won’t evolve. The future isn’t about ‘us vs. them’—it’s about humans and AI co-evolving together."

– Erica Orange

"AI isn’t just changing how we work—it’s changing what it means to be human. We are interlacing with technology more deeply than ever, and in the future, AI won’t just be something we use—it will be something we integrate into ourselves."

– Pedro Uria Recio

Alexandra Diening is Co-founder & Executive Chair of Human-AI Symbiosis Alliance. She has held a range of senior executive roles including as Global Head of Research & Insights at EPAM Systems. Through her career she has helped transform over 150 digital innovation ideas into products, brands, and business models that have attracted $120 million in funding . She holds a PhD in cyberpsychology, and is author of Decoding Empathy: An Executive's Blueprint for Building Human-Centric AI and A Strategy for Human-AI Symbiosis.

Mohammad Hossein Jarrahi is Associate Professor at the School of Information and Library Science at University of North Carolina at Chapel Hill. He has won numerous awards for teaching and his papers, including for his article “Artificial intelligence and the future of work: Human-AI symbiosis in organizational decision making.” His wide-ranging research spans many aspects of the social and organizational implications of information and communication technologies.

Erica Orange is a futurist, speaker, and author, and Executive Vice President and Chief Operating Officer of leading futurist consulting firm The Future Hunters. She has spoken at TEDx and keynoted over 250 conferences around the world, and been featured in news outlets including Wired, NPR, Time, Bloomberg, and CBS This Morning. Her book AI + The New Human Frontier: Reimagining the Future of Time, Trust + Truth is out in September 2024.

Pedro Uria-Recio is a highly experienced analytics and AI executive. He was until recently the Chief Analytics and AI Officer at True Corporation, Thailand’s leading telecom company, and is about to announce his next position. He is also author of the recently launched book Machines of Tomorrow: From AI Origins to Superintelligence & Posthumanity. He was previously a consultant at McKinsey and is on the Forbes Tech Council.

Websites:

Alexadra Diening

Mohammad Hossein Jarrahi

Erica Orange

Pedro Uria Recio

LinkedIn Profiles:

Alexandra Diening

Mohammad Hossein Jarrahi

Erica Orange

Pedro Uria Recio

Companies & Organizations

Books & Publications

Technical Terms & Concepts

Ross Dawson: So, you've recently established the Human-AI Symbiosis Alliance, and that sounds very, very interesting. But before we dig into that, I'd like to hear a bit of the backstory. How did you come to be on this journey?

Alexandra Diening: It's a long journey. I'll try to make it short and interesting.

I entered the world of AI almost two decades ago through a very unconventional path—neuroscience. I'm a neuroscientist by training, and my focus was on understanding how the brain works. Naturally, if you want to process all the neuroscience data, you can't do it alone. You inevitably have to touch upon AI. That was my gateway into the field.

As I started working with AI, I gained a basic understanding of how it operates from a technical perspective as a scientific discipline. At that time, there weren’t many people working in this kind of AI, so the industry naturally pulled me in. I started working in the business application of AI, progressively shifting from neuroscience to AI deployment within a business context. I worked with Fortune 500 companies across life sciences, retail, finance, and many more industries.

That was my entry—my "chapter one"—into the world of AI. But as I began deploying AI within real businesses, I started noticing patterns. Sometimes AI projects succeeded, and sometimes they failed. I realized that success was most often achieved when we doubled down on human-centricity. That was an easy concept for me to grasp because cognitive science is my foundation.

This human-centric approach became even more important with the emergence of generative AI. AI was no longer just in the background, crunching data and steering our decisions without us realizing it. AI has been around for quite some time, but suddenly, we could interact with it directly, almost like an agent. We could communicate with it using our language. It could capture emotions, build relationships with us, and augment our capabilities. It was no longer just a tool—it was becoming a social-technological actor.

This realization led us to our hypothesis: generative AI is the first technology with an almost natural, almost default propensity to form a symbiotic relationship with humans. It’s not just a tool that does something or doesn’t—it’s about mutual interaction.

The term "symbiosis" sounds very romantic, particularly because of the way pop culture has shaped our understanding of it. But in nature, symbiosis manifests across a spectrum of outcomes. It can be highly positive and mutualistic, where both parties benefit—humans improve, and AI gets better. However, it can also be parasitic, where one party benefits at the detriment of the other.

This pattern became clear to me, especially as generative AI adoption increased. I saw the emergence of what I call "parasitic AI," and that realization started stealing my sleep. I was no longer proud of the AI world we were building.

At the time, I was working for a multibillion-dollar tech company, and I doubled down on advocating for responsible AI and human-centric practices. But even with all the support in the world, I quickly realized that corporate agendas and business impediments limited the impact I could make. That’s why we established the Human-AI Symbiosis Alliance.

Our goal is twofold: first, to educate people that AI can be parasitic. It’s not just a happy story, and it’s not simply about AI taking over—it’s about how we deploy it. Second, we want to teach and empower companies to steer AI development away from parasitism and toward mutualistic AI.

Ross: We are deeply immersed in digital environments, and these systems are becoming increasingly human-like. You mentioned the idea of positive symbiosis. Achieving that requires well-designed systems and an understanding of how humans behave. What do you see as the foundational leverage points that can shift us toward a positive and constructive symbiosis between humans and AI?

Alexandra: The most important realization is that AI is not a living entity. It’s just a large dataset. It doesn’t have consciousness, intent, or agency. Instead of seeing AI as something that will inherently harm us, we need to take responsibility for how we deploy it.

Of course, we need to ensure AI is properly regulated, that it is trained on unbiased data, and that we establish appropriate guardrails. But there’s another chapter of the conversation that very few people talk about, and it keeps me up at night: the way we deploy AI.

Deploying AI in a way that doesn’t harm individuals or companies is critical. No company wants to build parasitic AI within its environment. The main issue in deployment comes from literacy. Many software engineering companies are now venturing into AI without realizing that AI development is fundamentally different from traditional software development.

You cannot deploy AI the same way you deploy web pages or apps. It has a completely different lifecycle, set of activities, and expertise requirements. Raising awareness about this difference is crucial.

Beyond that, we need frameworks—structured processes that guide responsible AI deployment. We also need to recognize that AI is not just a technology we implement; it’s a symbiotic relationship we must architect. That means not only enhancing employee efficiency in the short term but also ensuring that AI doesn’t erode human skills over time. Otherwise, we risk creating a workforce that is highly efficient but, in the long run, less capable.

Another crucial element is measurement. The traditional ways we measure technology success—primarily through productivity and efficiency—are outdated for AI. We need to consider additional factors, such as how AI impacts innovation, employee well-being, and a company's brand relationships. Instead of being shortsighted, we need a long-term focus on AI’s broader impact.

Finally, AI brings entirely new risks, many of which are unprecedented. A very personal and tragic example is the case of a teenager who took his own life after interacting with an AI chatbot.

When I used to warn clients about the importance of setting the right level of anthropomorphism and properly guarding AI to prevent harm, it often felt abstract. But now, unfortunately, we have a very tangible example of how things can go wrong.

The key takeaway is that building a responsible, mutualistic AI requires expertise, proper architectural planning, accurate measurement frameworks, and a heightened awareness of risks. If we get those things right, we can steer AI away from parasitism and toward a future where it genuinely benefits society.

Ross:In this section, we hear from Mohammad Hossein Jarrahi, Associate Professor at the University of North Carolina, Chapel Hill, from Episode 62.

Ross: So, you have been focusing on human-AI symbiosis. I’d love to hear how you came to believe this is where you should be focusing your energy and attention.

Mohammad Hossein Jarrahi: It was in 2017, and I was stuck in traffic. If I want to tell you the story, there was an IBM engineer being interviewed on NPR. They were asking him a bunch of questions about the future of AI.

This was before the rise of ChatGPT and what I would call the consumerization of AI. As I was sitting in traffic with not much to do, something clicked. The engineer was providing examples that fit into three categories: uncertainty, complexity, and eco-locality.

As soon as I got home, I immediately started sketching out an article and finished writing it within two weeks. The idea was that we, as humans, have very unique capabilities, but we tend to underestimate them. At the same time, the smart technologies we see today—at that time, primarily powered by deep learning—are inherently different from previous information technologies.

This means we need a completely different paradigm to understand how humans and AI can work together. AI isn’t going to make us extinct, but we shouldn’t treat it as just another infrastructure technology, like Skype or other traditional communication tools.

That’s when I realized that the term human-AI symbiosis—which comes from biology—was a perfect way to describe how two sources of intelligence can work together.

Ross: That concept is very much aligned with my work and the people I engage with. The key question is, how do we make it happen? There are quite a few people exploring this path, but we don’t yet have all the answers.

What are some of the pathways that could move us toward effective human-AI symbiosis?

Mohammad: It really depends on the context. That’s the crux of the issue I’ve been exploring in my articles.

The question of how much we can delegate to AI isn’t black and white. It exists on a spectrum between automation and augmentation. AI provides dual affordances—it can automate tasks or augment human capabilities.

Automation means AI performs tasks autonomously with minimal supervision. Augmentation, on the other hand, keeps humans deeply involved, making them more efficient and effective.

The balance between automation and augmentation depends on the context:

Machines are excellent at handling tasks that are repetitive, data-centric, and do not require intuition or emotional intelligence. However, humans excel at exception handling—making nuanced judgment calls.

For example, consider loan applications:

That’s why, when it comes to organizational decision-making, AI shouldn’t be the sole authority. Stakeholder interests are often in conflict—what benefits shareholders may harm employees or customers. AI tends to optimize for one metric, but a human leader must strike a balance among competing priorities.

Ross:I think a lot about the architecture of AI integration.

Keeping humans in the loop is important, but where should humans be involved? That depends on the organization, decision type, and context.

Are there structured ways we can design points of human involvement—whether in exceptions, approvals, or shaping judgment?

Mohammad: The simplest answer is that humans should be involved whenever intuition is required.

In my article on human-AI symbiosis, I described two decision-making styles:

For example, in algorithmic management, AI can assist managers, but the higher you go in an organization, the more important intuition becomes. Research in management and psychology has shown that holistic decision-making—which accounts for multiple stakeholders—relies heavily on intuition.

If AI only optimizes decisions based on data, it risks missing broader considerations, such as company culture, long-term brand impact, or ethical concerns. That’s why judgment calls must remain in human hands.

Ross: Next, we hear from Erica Orange, futurist and author of AI and the New Human Frontier, from Episode 59.

Ross: What will allow us to master AI and ensure it benefits humanity?

Erica Orange: That’s such a great question. I often talk about the difference between lifelong learning and lifelong forgetting.

It’s common to hear that we should all be lifelong learners—constantly acquiring new knowledge to stay relevant. But if we keep layering new information on top of outdated thinking, we won’t truly evolve.

We must also become lifelong forgetters—letting go of outdated assumptions, biases, and ways of working.

I often tell my clients and audiences to identify one or two things they’re holding onto that no longer serve them. It could be a belief, a work habit, or an outdated mental model. The faster we embrace forgetting, the more space we free up for new ways of thinking.

Another key point is to embrace the "AND" mindset instead of thinking in polarized extremes.

We live in a world of hyper-polarization—social media echo chambers and tribalism reinforce "us vs. them" thinking. But the future isn’t either-or—it’s about "and."

For example, when discussing humans and AI, there’s often fear of an "AI takeover." But AI isn’t replacing us—it’s collaborating with us. The reality is one of coexistence and co-evolution.

The same applies to progress and stagnation, chaos and creativity, imagination and inertia—these forces always exist together.

Ross: Finally, we hear from Pedro Uria Recio, author of Machines of Tomorrow, from Episode 50.

Pedro Uria Recio: In Machines of Tomorrow, I explore AI through human history.

From ancient aspirations of creating human-like machines to today’s generative AI revolution, AI has always been intertwined with our progress.

One of the book’s key concepts is interlacing—the idea that humans and AI will become more intimately connected.

Right now, we use smartphones for everything. The fact that they exist outside our bodies is merely an anecdote—in the future, they will be inside us.

Brain-computer interfaces, robotics, and AI-driven medicine will interlace humans and AI, potentially transforming humanity into a new species.

This shift won’t happen overnight, but AI will be central to our evolution.

Ross: That wraps up this episode. Thank you to all our guests for their incredible insights on human-AI symbiosis.

The post Human AI Symbiosis Compilation (AC Ep77) appeared first on Humans + AI.

– Rita McGrath

Rita McGrath is one of the world’s top experts on strategy and innovation. She is consistently ranked among the top 10 management thinkers globally and has earned the #1 award for strategy by Thinkers 50. She is Professor of Strategy at Columbia Business School, and Founder of the Rita McGrath Group and Valize LLC. Her books include The End of Competitive Advantage and Seeing Around Corners.

Website:

Rita McGrath

Valize

LinkedIn Profile:

Rita McGrath

University Profile:

Rita McGrath

People

Companies & Organizations

Books

Technical Terms & Concepts

Ross Dawson: Rita, it is fantastic to have you on the show.

Rita McGrath: Thank you very much for inviting me.

Ross: So my personal experience is that, over time, the world has come towards me, and what I’ve been thinking has become more and more of a reality. That strikes me very much with your work. I think you’ve been incredibly prescient. A lot of the themes you’ve worked on for years are even more relevant today than they were earlier. Has that been your feeling?

Rita: It has. It has. I mean, I was writing about what we would now recognize as the lean startup movement back in the ’90s. Clayton Christensen and I were working together on his idea of disruptive technology. My book The End of Competitive Advantage, which basically argued that competitive advantages last for shorter and shorter periods of time, came out in 2013, and people are still saying, "Wow, that’s so interesting." So it is that kind of feeling.

Ross: In particular, you’ve talked about transient advantage. A very long time ago, that advantage has become more and more transient, which we can frame as acceleration. And I think that there used to be a bit of debate—is the world accelerating, or is it just a feeling that it’s accelerating? So what’s your perception today in terms of where we might move forward, especially regarding the pace of change in business, strategy, and competitive advantage? Is this acceleration likely to continue?

Rita: Yes. To quote Ray Kurzweil, any system that embeds experience-based learning—trial and error learning—tends to follow an exponential change pattern. It’s not additional, it’s not linear—it’s exponential. And we, as human beings, experience that as things moving faster and faster.

So, day one, it’s two. Day two, it’s four. Day three, it’s eight. Eventually, these exponential curves take off, and I think we’re seeing quite a bit of that with the current developments in AI at the moment.

Ross: You’ve pointed to this theme of inflection points. How would you frame some of the current developments in AI or its impact on business around that theme? Are we living through an inflection point or a phase—or a series of them at the moment?

Rita: Yeah, I believe we are. And I would say that there are multiple levels of inflection points.

At a 30,000-foot level, if you think about the financial and social structures of capitalist systems, they go through these 50- to 70-year cycles each time a new technology emerges that dramatically changes our ability to do something.

Going all the way back to the 1700s and the original Industrial Revolution, what you see happening is what I define as an inflection point—something that creates a 10x shift in what’s possible. In the Industrial Revolution, labor was automated. Then we had the mass production era—cars, suburbs, petroleum-based economies—that has been coming to the end of its S curve of delivering prosperity and productivity.

The next wave is really this era of digitalization, which I would date to the early ’70s, with the microprocessor and the earliest digital technologies. What digitalization does is change what’s possible by a factor of 10. Some of the effects are quite surprising.

For example, one of them is dematerialization—taking things that used to require their own physical device or material and digitizing them. Songs are a great example.

Back in the day, you had to buy records, have a record player, and all the associated technology. Later, it was CDs, where you had to pay for 18 songs when you just wanted one. Today, we don’t even buy songs anymore—we stream them on demand on a device that doesn’t require physical input. It’s all digital.

Ross: One of the key themes is strategy. How do we do strategy in a world where acceleration is happening, where there are all these overlaying themes, and these 10x shifts across different dimensions of work?

This brings us to human capabilities. Humans have finite cognition, even though we have extraordinary capabilities far beyond anything else. And now, we have AI to augment, support, or complement us. I’d like to dive deeper into this, but to start, how do you frame human capabilities in strategic thinking today, and how are they complemented by AI?

Rita: Sure. Well, I think, as I mentioned, human brains think in linear terms. We think immediately in terms of, I have to get from here to there to avoid a predator, and back when we were evolving, that worked pretty well.

But we don’t handle exponential systems well because they appear small and insignificant—until suddenly they don’t. It’s this whole gradually, then suddenly idea.

What I argue is that we need to supplement what our brains can manage on their own, and this is where AI comes in.

What I’ve set up with companies is a series of what I call time zero events—signals that indicate a future inflection point is approaching. We don’t know exactly when, but then we work backward and ask, Before that happens, what conditions need to be in place?

AI makes the process of sensing signals much faster and more efficient. It supplements our brains by sorting through massive amounts of data, identifying patterns, and alerting us to relevant developments.

AI isn’t very good at imagination—it’s better at hallucinations than true creativity right now. But by blending human imagination with AI’s ability to process vast amounts of data, we can create powerful tools for strategy.

Ross: That’s fabulous. Rita, it’s fantastic to see the body of your work. And I think not just the open-mindedness, but also the questions you’ve asked, have anticipated the world we’re living in today. Your work is extraordinarily relevant.

So where can people go to find out more about your work?

Rita: Well, ritamcgrath.com is a good place to start. That’s my personal website, where you can find all kinds of information, downloadable articles, and so forth.

I also publish regularly on LinkedIn, Medium, and Substack, so you can find me in those places.

And for those who might be interested in more of an advisory touch, I have a sister company called Valize—that’s V-A-L-I-Z-E—that is figuring out what this new model for consulting is going to look like. I’m not sure we have the answer yet, but we’re certainly happy to go on the journey to figure it out.

Ross: Fantastic. Thank you so much for your time and insights, Rita.

Rita: Thanks.

The post Rita McGrath on inflection points, AI-enhanced strategy, memories of the future, and the future of professional services (AC Ep76) appeared first on Humans + AI.

– Christian Stadler

Christian Stadler is a professor of strategic management at Warwick Business School. He is author of Open Strategy, which was named as a Best Business Book by Financial Times and Strategy + Business and has been translated into 11 languages. His work has been featured in Harvard Business Review, New York Times, Wall Street Journal, CNN, BBC, and Al Jazeera, among others.

Website:

Christian Stadler

LinkedIn Profile:

Christian Stadler

University Profile:

Christian Stadler

Companies & Organizations

Books

Technical Terms & Concepts

Ross Dawson: Christian, it's a delight to have you on the show.

Christian Stadler: Thanks for having me, Ross. It is a delight for me as well.

Ross: So, you have been delving deep into a lot of your background in open strategy. You've also been looking at the role of AI in strategy and strategic decision-making. At a high level, how do you see the role of AI in strategy making today?

Christian: I’m an optimist. I think generally, by nature, and also when it comes to how AI can actually be useful for strategists, more and more people are coming to see AI as a partner in many different areas of what we do. I think that's true for strategy as well. We have some form of co-decision-making, co-intelligence, or an additional voice that we can use in the strategy-making process. For that, it’s really cool.

Ross: These are human-first processes, I suppose. The more complex the decisions are, the more multifaceted they become, and the more the human element needs to be at the forefront. Strategy seems to fall into that category. What are the places where AI might provide support, complementary perspectives, or analysis that are particularly valuable?

Christian: Strategy, obviously, consists of different "boxes" or activities. Some involve coming up with new ideas—something new you want to do in your strategy. Other parts involve fine-tuning and formulating the strategy. Then there’s the execution and implementation side. Probably in each of these aspects, it makes sense to use AI in slightly different ways.

When it comes to ideation, I can ask a tool for ideas, such as setting up a new product line. I played around with this early on when ChatGPT started gaining traction. Even then, it was phenomenally good if you guided the conversation as a strategist. If you just ask ChatGPT, you get generic suggestions, and sometimes they don’t make sense. For example, I once asked for a suggestion for a streaming service. One idea was to create some form of entertainment platform and partner with universities.

Being a professor, I know that universities don’t work like that. Professors aren’t told to participate by some central directive. You need to find ways to motivate individual professors. As I pushed the platform further, better ideas came up. As long as you drive the conversation and are smart about it, AI can provide good ideas.

When it comes to fine-tuning and formulating, the tool can be quick. I’ve been experimenting with a company in Austria for over a year. They make sneakers—Gieswein. We tried seeing what happens in board meetings when we bring ChatGPT into the mix. For instance, when we needed a press release, the tool quickly drafted something. In this case, we didn’t need an agency, which saved time and resources.

However, when it comes to execution, that’s more of a human game. You need to convince people to buy into ideas and feel comfortable with new directions. AI has limitations here, but other tools can help involve more people. Greater involvement aids implementation.

Ross: There’s a lot there I’d like to dig into. We might do a bit of hopping around.

Christian: It’s a bit long-winded, isn’t it? I just keep talking on and on. My bad.

Ross: It’s all good. One interesting point is that part of Amazon’s internal processes involves starting with a press release for a potential product. Then they work backward to figure out how to achieve it. That’s something ChatGPT can facilitate in board meetings. You can draft a press release and discuss if this is something you want to pursue.

Digging into the boardroom specifically, how do you see AI being valuable when working with a group of directors? For instance, red-teaming—having the AI critique decisions—seems promising. What are other potential applications?

Christian: One surprising and insightful aspect of using ChatGPT in boardrooms was its disruptive nature. Imagine a group that has worked together for a long time. The process is smooth because they know how each other thinks. Then ChatGPT comes in and disrupts the flow. For instance, I might ask someone to read a page of suggestions from ChatGPT mid-meeting. It forces the group to stop and think.

Initially, I thought they would hate it, but they actually liked it. The disruption brought up ideas that wouldn’t have otherwise come up in the discussion. It changed the process in a positive way, rather than simply adding information.

Ross: So you were distilling conversations, summarizing, and presenting them to the board?

Christian: Essentially, I was a disturbance. For example, when discussing market entry into the U.S., I’d interrupt and say, “Here’s what ChatGPT suggests.” Having to read and discuss AI-generated content mid-meeting isn’t smooth, but that lack of smoothness was beneficial.

Ross: That reinforces the idea that the facilitator plays a critical role. You acted as an AI-enabled facilitator. Your choice of interventions determined the success of the process.

Christian: Absolutely. We tried different approaches: preparing content beforehand, engaging during the meeting, and doing post-meeting analysis. When the tool worked independently, the output was too superficial. It needed human direction. I’m not an industry expert, but with a feel for strategy, you can create significant benefits.

Ross: One of the specific applications of AI is strategic decision-making, where you already know what decision needs to be made. The decision-making process typically involves defining the decision, generating options, assessing those options, and ultimately making a choice. AI can assist with ideation and evaluating options.

How do you see AI's role evolving in formal strategic decision-making processes, both today and in the future?

Christian: You mentioned options, and I’ve always been a big fan of scenario planning—drawing pictures of what the future could look like in various versions. Some companies use two scenarios; others prefer four, making it more complex. Whatever strategy you pursue, it needs to be tested against these different futures.

AI tools like ChatGPT are excellent at generating plausible future scenarios. Of course, human direction is necessary, but AI can support the process. Writing compelling, coherent stories about the future is a skill not everyone possesses, and AI can facilitate this.

For now, I see AI primarily as a facilitator. In the medium term, it helps strategists think through different possibilities. Whether AI will ever be capable of independent strategic thinking, I can’t say—I’m no magician. But for now, its power lies in augmenting human intelligence rather than replacing it.

Ross: In some of your work, you’ve referenced ethical concerns. Humans have the ability to grasp broader context, values, and the human experience in ways AI cannot. Personally, I believe AI will remain a strong supporting tool, but I doubt it will take over higher-order strategic decision-making.

Christian: I agree. As you know, I’ve long advocated for involving more people in strategy-making. Opening up the process brings in fresh, unconventional voices, leading to better strategies. AI can serve as one of those voices—offering unexpected insights that force us to think differently.

However, human context is essential. AI-generated suggestions must be assessed within the company’s reality. AI becomes even more valuable when integrated with internal company data, allowing for more tailored insights. That said, I’m cautious about assuming all relevant data is neatly captured. In big tech, this might be the case, but for many medium-sized businesses, it's not.

For example, I spoke with a CEO who runs a company that makes high-end ski gloves. His strategic decisions—what products to produce and in what quantities—aren’t based on hard data. Instead, he relies on conversations with retailers and industry experts. This highlights a limitation of AI: in many cases, businesses don’t have the vast datasets AI needs to be truly effective.

Ross: Let’s dig into open strategy. Could you provide a simple framing of what open strategy is? And how does it connect to AI?

Christian: The easiest way to understand open strategy is to contrast it with traditional strategy-making. Historically, strategy was developed behind closed doors by a small group—perhaps with the help of a consulting firm.

Open strategy, on the other hand, involves bringing in more voices. This approach not only generates fresher, better ideas but also makes execution smoother. The majority of failed strategies don’t fail because they were bad ideas—they fail due to poor execution. Various surveys suggest that up to 90% of strategic failures stem from execution issues.

When people are involved in strategy-making, they develop buy-in. They also begin to see how strategic goals connect to their work. In our book Open Strategy, we surveyed executives who had implemented open strategy. About 69% said it led to better ideas, and 70% noted that execution was significantly improved.

Ross: We can think of open strategy in different layers. One layer involves opening strategy within the organization, allowing all employees to participate. Another layer involves engaging external stakeholders—partners, suppliers, customers, or even the public.

What are your thoughts on these different levels of openness?

Christian: Absolutely. There are different degrees of openness. You don’t necessarily have to involve all employees—you might just expand participation beyond the usual small group. This is the most common approach and brings significant benefits, even in hierarchical organizations.

For example, I worked with a company in the Middle East, where hierarchy is deeply ingrained. Initially, there was hesitation about involving middle management in strategy-making. Eventually, they agreed, and the results were fantastic. It helped align the organization behind the new strategy.

In this case, we first collected middle management’s input separately because they might have hesitated to speak openly in front of top executives. Later, we shared their insights with leadership. This process built enough trust that in a subsequent round, both groups could participate together.

As for external engagement, it depends on the phase of strategy-making. During the ideation phase, involving external voices can be valuable. You don’t need to share company secrets—just frame the challenge broadly and let external contributors provide fresh ideas.

Even the U.S. military has done this. The Pentagon has held open exercises where the public contributes strategic insights, but they don’t necessarily disclose how those insights are used.

For execution, however, you want broader internal involvement. Everyone in the company needs to understand what’s happening. IBM ran one of the largest open strategy initiatives, involving 160,000 participants. Managing a discussion at that scale requires AI-powered tools to structure and synthesize input.

Ross: Open strategy can be seen as a form of collective intelligence. Whether it’s eight board members, 100 managers, or an entire organization, the challenge is structuring participation effectively.

What’s the state of the art in integrating diverse perspectives into a coherent strategy? How can we improve?

Christian: I have to admit, I’m still a bit old-school when it comes to strategy. I prefer in-person workshops because they allow for deeper discussions. That said, large-scale engagements require digital tools.

A structured approach is key. One method is to start with a broad survey to identify major trends. This helps leadership gauge the organization’s sentiment. Understanding what people think is happening is just as important as knowing what’s actually happening. If leadership’s actions contradict employees’ perceptions, it can create resistance.

Next, bring people into structured workshops. Different teams can develop and pitch ideas. A “Dragon’s Den” format works well—teams compete to refine the best ideas. Facilitators play a crucial role in guiding discussions and ensuring productive outcomes.

Ultimately, open strategy isn’t about turning companies into democracies where everyone votes on decisions. Instead, leadership uses the insights generated through participation to make informed choices. The key is communicating back to employees—explaining what decisions were made and why. People don’t expect to be the final decision-makers, but they value having a voice.

Ross: In a world of accelerating change—particularly with AI—leaders need to refine new capabilities.

What skills do senior executives, board members, and strategy-makers need to be effective in today’s landscape?

Christian: First, they need to engage with AI. It’s as simple as replacing Google with an AI tool when searching for information. Play around with it, get familiar. These models are user-friendly, and you don’t need programming skills to experiment.

Second, leaders must navigate the increasing entanglement between business and politics. In past decades, it was easy to overlook politics, but that illusion is gone. Leaders must understand how to operate in politically charged environments.

It’s not about whether a company is conservative or liberal—successful brands exist at both extremes. Nike is known for progressive values, while Chick-fil-A is conservative, yet both maintain broad customer bases. The problem arises when companies appear inconsistent or opportunistic. For example, many big tech firms once positioned themselves as liberal but later courted the Trump administration. That shift alienated both sides.

Finally, attracting and retaining talent is critical. Many CEOs cite talent shortages as their top concern. Engaging employees, making work meaningful, and fostering motivation will be key leadership skills.

Ross: That ties into the aging workforce issue. Countries with declining populations are often the most receptive to AI and robotics. This shifts the workforce balance, making talent attraction even more crucial.

Christian: Absolutely. Immigration could help, but politically, it’s difficult. Interestingly, even Elon Musk—despite his conservative shift—supports H-1B visas because he recognizes the need for skilled immigrants.

Ross: As we wrap up, what excites you most about your upcoming research, particularly regarding AI and strategy?

Christian: I’m exploring the intersection of corporate strategy and individual decision-making. I’m also using AI to analyze emotions in strategic discussions. For example, I’m working with IBM data to study how emotions impact idea adoption in large-scale strategy sessions. Using AI for research is something I really enjoy.

Ross: That sounds fascinating. Thank you for sharing your insights!

Christian: Thank you! It was a pleasure.

The post Christian Stadler on AI in strategy, open strategy, AI in the boardroom, and capabilities for strategy (AC Ep75) appeared first on Humans + AI.