One hot Texas afternoon, four people climbed into a car with no air conditioning and drove 85 km to Abilene for dinner. The drive was brutal. The food was bad. They came home exhausted.

Not one of them had wanted to go.

THE STORY BEHIND THE PARADOX

Jerry B. Harvey, professor of management science at George Washington University, told this story about his own family in his 1974 paper "The Abilene Paradox: The Management of Agreement."

His father-in-law suggested the trip. His wife said it sounded great. Everyone agreed. Only on the way back, in silence, did the truth come out—starting with his mother-in-law: "To tell the truth, I didn't enjoy it. I only went because the three of you were so enthusiastic."

Every single person had gone along because they thought the others wanted to.

THE COUNTERINTUITIVE CORE

After several episodes on managing conflict, Harvey flips the question.

"It's not about managing conflict better. The more dangerous organizational failure is the inability to manage agreement."

Groups take actions that contradict what every individual privately wants—because each person wrongly believes everyone else is enthusiastic. The result is a decision nobody supports, frustration all round, and bewilderment about how it happened.

The mechanics: members privately agree about the situation and what to do, but fail to communicate it. Everyone misreads the collective reality, acts against their own wishes, and then blames each other—which sets up the next trip to Abilene.

WHY IT'S NOT GROUPTHINK

A crucial distinction. In groupthink, people genuinely talk themselves into believing the bad decision is right. In the Abilene Paradox, they never believe it—they just go along.

The engine is the fear of separation: the anxiety that voicing dissent will get you ostracized. People run "negative fantasies" about what happens if they speak up, and the imagined risk of exclusion outweighs the real cost of going along.

"Clark's first stage of psychological safety is inclusion safety. Here we have the opposite—the fear of exclusion."

WHY THE LEADER SHOULD SPEAK LAST

A practical tip: in a meeting where you want honest input, be the last to speak. The moment you reveal your view, people read your face and align to it. The hope of belonging quietly overrides their real opinion.

THE JAKOB SCENARIO ON ROLEPLAYS.AI

This is exactly why Bernhard built a scenario to practise it. "Where Are We Actually Going?"—free in June.

Five friends from their London Business School MBA plan a trip. It opens in a WhatsApp group. Jakob says: "Let's go to Greece." Jonah: "Greece is nice, I'm easy though." Another: "Croatia would be lovely…" And it drifts. Your job: get the group to a place everyone actually wants. Harder than it sounds. (Make all the personas agree on one location and you might win a prize.)

"Reading about the Abilene Paradox is the easy part. Practice is where it shows if we actually inhaled what we learned."

The scenario isn't negotiation or facilitation. It's closer to hearing what people are not saying—and slowing the group down long enough to find out what they actually want before the booking link goes out.

ONE TOOL TO TRY

The "three hats" exercise (adapted from DreamWorks): assign an optimist, a realist, and a pessimist, and argue the case from all three roles. Suddenly you're not the person saying "I don't want to go to Abilene"—you talk it through, see the pros and cons, then decide.

Today's invitation: start listening for what people don't say but actually think.

REFERENCE: Harvey, J. B. (1974). The Abilene Paradox: The Management of Agreement. Organizational Dynamics.

LINKS: bernhardkerres.com | roleplays.ai

#AbileneParadox #PsychologicalSafety #Teams #Leadership #Coaching

Last week Bernhard claimed you don't need to be friends to be a high-performing team. Then he found a study that, at first read, said he was wrong. Was he?

THE HARDEST POSSIBLE TEST

A string quartet is the purest intense work group that exists. Four people, completely interdependent, who cannot produce a single bar alone. They rehearse six hours a day, tour together, sit beside the same three people for years—one quartet in the study, for 34 years. The musicians described it as a marriage—not to one person, but to three.

If friendship matters anywhere, it matters here.

Murnighan and Conlon (Administrative Science Quarterly, 1991) studied 20 of Britain's professional quartets, interviewing 80 musicians. Nine years later, they checked who survived. Over half had folded.

THE COMPLICATION—AND THE REFRAME

The successful quartets did tend to be friends. So on the surface, the study contradicts last week's claim.

But dig into why. They weren't held together by friendship—they were held together by an obsession with the music. Members of the best groups independently, almost unanimously, called it exactly that.

"The friendship was not the engine. The shared, non-negotiable commitment to excellence was the engine. Friendship grew out of it."

It's not liking each other that produces performance—it's a shared, obsessive commitment to the same standard.

THREE TENSIONS EVERY TEAM LIVES WITH

Three paradoxes no quartet can resolve. The successful ones managed them quietly; the unsuccessful ones tried to fix them and broke.

1. Leader vs. Democracy. The music gives the lead to the first violinist, but every player joined for an equal voice. The successful first violinists led totally—and sincerely advocated democracy at the same time. The researchers called it "effective inconsistency." The lesson: pretending the hidden hierarchy isn't there breaks groups.

2. The Second Fiddle. The second violinist is often more technically skilled than the first—the parts are harder—yet gets none of the acclaim. A famous second's metaphor: a quartet is a bottle of wine. The first violin is the label, the cellist is the bottle, and the second violin and viola are the wine—the actual contents. Successful quartets made their second feel essential; failing ones treated them as a lesser first.

3. Confrontation vs. Compromise. The successful quartets did not resolve conflict through open confrontation. Their saying: "Either we play or we fight." They let trivial disagreements dissolve; the ones that mattered worked themselves out in the playing. This complicates challenger safety—and refines it. It isn't challenging everything all the time; it's building enough safety that you can challenge, then choosing what's worth it.

THE DEEPER PATTERN

The standard advice was to confront your paradoxes openly. The data said something truer: the best quartets recognised their contradictions and did not try to untangle them. We're trained to resolve, fix, align—but the highest-performing teams develop the maturity to live inside unresolved tension. And they worked through disagreement not by talking it to death, but by playing. The thesis behind the Rehearsal Gap and RolePlays.ai: you resolve the hard things in the doing, not the discussion.

THE CORRECTION

"You don't need to love the people you work with. You need to be obsessed, together, with the same excellence. The friendship, if it comes, comes after."

Ask of your own team: do we share a drive for excellence, or are we just trying to get along? That's the difference between a quartet that lasts 34 years and one that folds.

REFERENCE: Murnighan & Conlon (1991), The Dynamics of Intense Work Groups: A Study of British String Quartets. Administrative Science Quarterly, 36(2).

LINKS: bernhardkerres.com | roleplays.ai

#PsychologicalSafety #Teams #Leadership #Coaching #HighPerformance

Flutist Agnes Vass asked Bernhard a question: what if you applied the Fifth Stage of psychological safety to an orchestra?

It turns out an orchestra is one of the best models we have for how high-performing teams work.

(With thanks to Agnes Vass, co-principal flute at the Bremerhaven Philharmonic and founder of the Body Mind Music Lab.)

WHY ORCHESTRAS?

Peter Drucker used orchestras constantly as a model for organisations. So has Bernhard, given his background.

The reason: an orchestra of 70 to 150 musicians creates an outstanding performance in three days—often led by a conductor they have never met before. The analogies for management write themselves.---

DUNBAR'S NUMBER

British anthropologist Robin Dunbar found that humans can maintain around 150 meaningful relationships at once (his average: 148.6).

The proof arrived with social media. Remember the early joy of Facebook—reconnecting with old friends? Then somewhere around 150–200, your feed filled with people you didn't really care about. Dunbar's number, demonstrated at scale.

Orchestras sit right inside that number. So does W.L. Gore, the company behind Gore-Tex—they cap units at around 200 people and build a new site rather than exceed it. The belief: people at work should genuinely know each other.

THE REFRAME: RELATIONAL SAFETY ISN'T FRIENDSHIP

Are 140 orchestra members all good friends? No. Some are close. Some can't stand each other. And yet the best orchestras deliver extraordinary performances.

"The safety a good orchestra has is this: even if I don't like my colleague, I know they are committed to the highest performance, just as much as I am."

Relational safety, properly understood, is built on a shared, explicit common goal—the same standard of quality, the same drive, the same dedication to practice. Not affection.

"If you're a flutist and you haven't practiced, every person in the audience will hear it."

FUZZY GOALS vs. MOTIVATIONAL GOALS

Organisations often run on fuzzy goals—"increase turnover by 10%." That's like telling an orchestra to finish five minutes early by playing faster. Nobody is moved by it.

Motivational goals are about a meaningful outcome: electrifying the audience, leaving the customer completely wowed. When everyone is committed to it, everything changes.

Underneath it: a commitment to practice. Musicians practice. Most managers wing it 80% of the time. That's why Bernhard built RolePlays.ai—a place for leaders to practice the difficult conversations.

DIVERSITY: HACKMAN'S ORCHESTRA RESEARCH

In the 1980s, J. Richard Hackman of Harvard studied women in orchestras. At the time, many were all-male—the Vienna Philharmonic didn't admit women until a US tour forced the change.

What Hackman found:

Below 10% women: high turnover, mobbing, sexism. Women leave.

Between 10% and 33%: a hard struggle.

Around 33%: an equilibrium. Men and women playing together become natural. Sexism drops. Performance improves. Women stay. You can even hear it—diversity changes the sound.

Not only a values argument: listed companies with diverse boards significantly outperform all-male ones.

THE THREE REQUIREMENTS FOR THE GROWTH ZONE

To bring a team into the growth zone—where breakthroughs happen—you need three things beyond relational safety: a shared, motivating sense of purpose; a genuine commitment to practice; and space for diversity.

True for a 140-person orchestra. Equally true for a team of five.

Jon Katzenbach put it well: what separates a high-performing team from a merely good one is that its members are committed to their own learning—and to each other's.

REFERENCES:

Dunbar, R. How Many Friends Does One Person Need?

Hackman, J. R. Leading Teams.Katzenbach, J. R. The Wisdom of Teams.

Agnes Vass — Body Mind Music Lab (Instagram).

LINKS: bernhardkerres.com | roleplays.ai

#PsychologicalSafety #Orchestra #Leadership #Teams #Diversity

Last week: Timothy Clark's four stages of psychological safety. This week: what they're missing.

Most teams that believe they have psychological safety are stuck in a place where everything looks safe but nothing moves.

TWO DIMENSIONS, NOT ONE

Psychological safety has two distinct dimensions. Most frameworks address only one.

Relational safety: How safe do team members feel with each other? Can they be candid, vulnerable, direct?

Environmental safety: How safe does the environment feel? Familiar room, known routines, predictable structures.

These two axes create four zones.

ZONE 1 — THE DANGER ZONE (low / low)

Toxic environment, hostile colleagues, chaotic, no trust. Individual: leave. Leader: act now.

ZONE 2 — THE PSEUDO SAFETY ZONE (high environmental / low relational)

The trap most "psychologically safe" teams are actually in.

Everything looks fine. Names on doors. Clear meeting structures. Everyone knows the systems. But nobody addresses conflict, nobody has the hard conversations, nobody pushes.

"It's an area where we feel okay, but nothing gets done in a way that moves the needle. Most people have already signed out—within themselves."

ZONE 3 — THE CRUISING ZONE (high / high)

Clear structure, trusted colleagues, hard conversations possible, listening genuine. Aim to be here 60–70% of the time.

But the big steps don't happen here.

ZONE 4 — THE GROWTH ZONE: THE FIFTH STAGE (low environmental / high relational)

Breakthroughs only happen when you deliberately remove environmental safety—and the relational safety is strong enough to hold the team together without it.

Warren Bennis, in Organizing Genius, showed this with case after case. Breakthrough innovation happens when small teams of people who trust each other are taken outside the big company and thrown into a garage. They no longer care about environmental safety. They are too deep in the work.

When Steve Jobs returned to Apple, he took his most trusted engineers to a different building and raised a pirate flag. Inside the original Macintosh, the engineers' signatures are moulded into the metal—where nobody could see them. They didn't need anyone to.

HIGH-PERFORMING TEAMS DON'T LAST

Jon Katzenbach (The Wisdom of Teams): high-performing teams exist for days, weeks, sometimes months. Never years.

"I ask leadership teams: what kind of team are you? They say: a high-performing team. I say: for how long? And what for?"

The Fifth Stage is the same. You go there for the breakthrough, then return to the Cruising Zone to recover. Then, if needed, go again.

THE TRAINING INTERVENTION

A five-day leadership course for consultants. The first two days deliberately build both kinds of safety—familiar room, breakout structures, personal stories, shared meals.

Wednesday morning: participants arrive to find the room destroyed. Everything overturned. Death metal at full volume.

"Most teams aren't ready. They run around like headless chickens. Some set up a desk and start working alone. Some go for another coffee. Rarely does the whole team come together, clean up, and start the morning themselves."

The exercise tests one thing: when environmental safety is gone, is relational safety strong enough to hold?

THE TAKEAWAY

No breakthrough happens in the same environment a team has always worked in. To unlock innovation, transformation, or change, environmental safety has to come down—but only after relational safety has been deliberately built.

Build the Cruising Zone first. Visit the Fifth Stage when the breakthrough is required. Return.

REFERENCES:

Clark, T. R. The 4 Stages of Psychological Safety.Edmondson, A. The Fearless Organization.Bennis, W. Organizing Genius.Katzenbach, J. R. The Wisdom of Teams.

LINKS: bernhardkerres.com | roleplays.ai

#FifthStage #PsychologicalSafety #Leadership #Teams #Coaching

So much has been written about psychological safety. This is a different angle—practical, tool-driven, and building toward something new.

Amy Edmondson gave us the concept. Timothy Clark made it practical with four stages. Bernhard walks through all four—and the tools to actually build them—before introducing his own fifth stage next week.

STAGE 1 — INCLUSION SAFETY

Do people feel they belong? It sounds trivial. It isn't.

"At a football match, they separate the fans into zones—because there is no inclusion safety. In leadership trainings, I look at the shoes. Fifty percent wear white sneakers. There's the white sneakers club."

Inclusion shows up in signs, clothing, and above all language. Bernhard describes a dinner where a comment meant to support a female colleague backfired—because it framed her as "female versus us" rather than simply one of the team.

STAGE 2 — LEARNER SAFETY

Can people admit they don't understand?

"I teach at universities, and regularly I find out by accident that people didn't understand the concept—because they didn't dare ask."

In high-charged groups—senior executives, consultants—nobody admits confusion. The leader's job: build a feedback loop so you know whether people actually learned, and make asking a question feel essential rather than embarrassing.

STAGE 3 — CONTRIBUTOR SAFETY

Can people contribute—and are they actually heard?

"From hundreds of coaching sessions: women in male-dominated teams are normally not listened to. The strangest thing? Their feedback says 'you should contribute more.' But they do. Nobody listens."

Two tools:

First, before discussion, give everyone two minutes to think. The fast talkers drown out the careful thinkers otherwise. Set a timer, then go around the table—and as leader, speak last. (John Katzenbach's wisdom of teams: the moment you speak, everyone aligns to you.)

Second, after the first round, each person must build on the previous person's idea—not add their own. It forces real listening and takes the conversation deeper.

STAGE 4 — CHALLENGER SAFETY

Can people challenge you—and you not take it personally?

"When senior executives say 'I welcome the challenge'—boy, they don't. Challenger safety is one of the hardest levels to reach, because so many of us take a challenge personally."

The tool: propose an idea, then go around the table—everyone must challenge it. No exceptions, even if you love it.

For bigger groups, an adapted Disney method: split into visionaries, realists, and pessimists. Each subgroup challenges every idea from their assigned role, regardless of what they personally think. Then rotate. The role gives people permission to challenge—and the idea gets dramatically better.

THE THROUGH-LINE:

"We can't take the four stages for granted. We need tools to build and grow them. Use them—don't assume they're already there."

NEXT WEEK — THE FIFTH STAGE:

"I'll take this a step further to the fifth stage—a concept I came up with over the last couple of weeks."

REFERENCES:

Edmondson, A. The Fearless Organization.Clark, T. R. The 4 Stages of Psychological Safety.Katzenbach, J. & Smith, D. The Wisdom of Teams.

LINKS: bernhardkerres.com | roleplays.ai

#PsychologicalSafety #Leadership #Teams #Coaching #Edmondson

What should I wear today? Coffee or tea? Espresso or cappuccino? Work from home or go to the office?

Our days are filled with questions.

"And one observation I made—with my clients, but also with myself—was that I increasingly use AI chatbots to get a decision. So I don't need to think."

THE STORY OF CROESUS:

Sixth century BC. The kingdom of Lydia. King Croesus, wealthy enough that 2,500 years later we still say "rich as Croesus."

The Persian Empire is rising on his border. Should he attack first, or wait?

Croesus tests every famous oracle. He gives every emissary the same instructions: on a specific day, ask what the king is doing. That day, Croesus does something deliberately bizarre—boils a tortoise and a lamb together in a bronze cauldron with a bronze lid.

Only Delphi gets it right.

He sends extraordinary gifts and asks his real question: should I attack the Persians?

The Oracle answers: "If you cross the river Halys, a great empire will be destroyed."

Croesus hears what he wants to hear. He marches. He is destroyed.

The great empire that fell was his own.

"The Oracle was not wrong. The Oracle was useless. Because Croesus had not gone to Delphi to think. He had gone to Delphi for an answer. And once he had one, he stopped thinking."

FRAME 1 — INTOLERANCE OF UNCERTAINTY (Carleton et al., 2007)

Nick Carleton runs the anxiety lab at the University of Regina. He works with people who deal with uncertainty for a living—police, paramedics, firefighters.

Intolerance of uncertainty is a deep, dispositional trait. Not pessimism. The inability to tolerate not knowing.

One scale item captures it: "I'd rather know bad news than stay in a state of uncertainty."

Carleton calls it transdiagnostic—it shows up across nearly every anxiety disorder. He has argued it may be the fundamental fear, evolutionarily ancient.

"It explains why we cannot leave the question alone. Why we will pay almost any price for an answer. Even the wrong one."

FRAME 2 — COMPUTERS ARE SOCIAL ACTORS (Reeves & Nass, 1996)

Two Stanford professors. The Media Equation. Clifford Nass died in 2013, age 55.

Random flattery from a computer made participants rate the experience more positively—even when told the praise was random.

Participants evaluated computers more favourably when they did the evaluation on the same computer. They were polite. To the computer's face.

When asked afterwards if computers have feelings, every participant said no. But they had behaved as if the computers were people. Mindlessly.

"When ChatGPT says 'I think this might be a good approach,' it triggers social scripts millions of years older than computers. We are trusting it before we have decided to."

"The Oracle at Delphi was a person speaking on behalf of a god. The Oracle on our laptop is a system speaking on behalf of nothing."

FRAME 3 — ARTIFICIAL EPISTEMIC AUTHORITIES (Hauswald, 2025)

Rico Hauswald, philosopher at TU Dresden. Social Epistemology, 2025.

Human authorities—doctors, scientists, judges—are accountable. When they're wrong, the wrongness is visible.

Benjamin Lange (Cambridge, 2025) names what's missing: AI lacks "epistemic failure markers."

"When ChatGPT is wrong, it sounds exactly the same as when ChatGPT is right. The voice is the same. The confidence is the same. There is no signal."

THE CLOSE:

"Croesus knew his answer came from a source whose history could be tracked. We have a faster oracle, a more available oracle, a more confident-sounding oracle."

"Whether we have a wiser one is a different question entirely."

REFERENCES:

Herodotus, Histories, Book 1.Carleton, R. N. et al. (2007). J. Anxiety Disorders, 21.Reeves, B. & Nass, C. (1996). The Media Equation.Hauswald, R. (2025). Social Epistemology, 39(6).Lange, B. (2025). Epistemic Deference to AI.

LINKS: bernhardkerres.com | roleplays.ai

#AI #Coaching #Croesus #ChatGPT #Leadership

You prompt an AI agent. The spinner starts. 30 seconds. A minute. Three minutes.

What do you do in that time?

"To be honest, I switch to another task. Then a third. Then I check email. By the time the agent finishes, I've forgotten why I prompted it. It's a nightmare."

FRAME 1 — CSIKSZENTMIHALYI: ATTENTION AS PSYCHIC ENERGY

Mihaly Csikszentmihalyi. Flow (1990). Attention is psychic energy—finite, that we can direct.

Flow: Psychic energy invested in a clear goal that matches your skills. Musicians know it on stage. Athletes know it in a race. Coaches know it in the room with a client.

Psychic entropy: When attention has nothing to focus on, the mind scatters. Rabbit-hole thinking, worry, regret, cyclical thinking.

AI wait times, by design, are periods of unfocused attention. Csikszentmihalyi predicted in 1990 what would happen if we built systems that created those periods at scale.

FRAME 2 — GLORIA MARK: THE 47-SECOND SCREEN

UC Irvine. 20 years tracking how long humans stay on one screen.

2004: 2.5 minutes. 2012: 75 seconds. 2016 onwards: 47 seconds.

"When external interruptions are removed, self-interruption spikes. We have trained ourselves to be pinged—so we ping ourselves."

After a single interruption: 23 minutes to fully refocus.

"The agent isn't the distraction. The agent creates the gap we fill with distraction."

FRAME 3 — SOPHIE LEROY: ATTENTION RESIDUE

University of Washington Bothell, 2009. When you switch from Task A to Task B, part of your cognitive activity remains stuck on Task A.

Residue is strongest when Task A was unfinished or time-pressured.

"By the end of the day, we've been working through layers of mental residue all afternoon."

This is not AI's fault. AI is doing what we asked. The question is what WE do while it works.

SOLUTION 1 — THE READY-TO-RESUME PLAN

Leroy & Glomb (2018). Tested across four studies.

Before launching an agent, take 30 seconds to write down: what you just completed, what's left, the first thing to begin when you return.

That closes the loop. Brain lets go. Agent works. You step away.

SOLUTION 2 — THE 40-SECOND WINDOW

A 2015 study: 40 seconds of looking at greenery measurably improved sustained attention.

This connects to Attention Restoration Theory (Kaplan, University of Michigan). Nature engages "soft fascination"—interest without effort. The brain recovers.

40 seconds of email or social media does the opposite: fatigue, overwhelm, reduced focus.

"Same 40 seconds. Two completely different outcomes. Stop trying to be productive during AI wait times. Look out the window. The science says it works."

SOLUTION 3 — BATCH YOUR AGENT WORK

Cal Newport, Deep Work. High-Quality Work = (Time Spent) × (Intensity of Focus).

Don't sprinkle agents through the day. Run them in cognitive batches. Same context, no switching cost.

THE PRACTITIONER'S RETURN TO PENCIL AND PAPER

"I built RolePlays.ai. I use Claude as my private learning university. And I return more and more to pencil and paper."

"Writing by hand is psychic energy directed into a single-channel activity. Flow architecture. The opposite of psychic entropy."

A creative day every Monday. A creative week every two months. That's how RolePlays.ai got built.

THE 12-WEEK PACT:

Next time the spinner starts—don't open email. Don't pick up your phone. Don't start another agent.

Look out the window for 40 seconds. Or write one line on paper.

Notice what happens to your mind.

"Are you in with me?"

REFERENCES:

Csikszentmihalyi, M. (1990). Flow.Mark, G. (2023). Attention Span.Leroy, S. (2009). OBHDP, 109(2).Leroy, S. & Glomb, T. M. (2018). Organization Science, 29(3).Lee, K. E. et al. (2015). J. Environmental Psychology, 42.Newport, C. (2016). Deep Work.

LINKS:www.bernhardkerres.com | www.roleplays.ai

#Focus #DeepWork #AI #Flow #AttentionResidue #Coaching

Can artificial intelligence bring us closer together as humans?

It sounds contradictory. But last week in Budapest, 18 young consultants from an international strategy firm proved that it can.

THE SETTING:

Advanced consulting skills training. 18 consultants, late 20s, two years experience. They know analysis, structuring, presentations.

"But what does really matter in consulting, especially in today's world? Where Claude and ChatGPT can do your analysis. And to be honest, what comes back is pretty good."

What's left? The ability to interact with clients in a way that goes beyond ticking boxes.

THE CASE — VOLLPENSION:

If you've been to Vienna, you might know Vollpension. A coffee house where grannies and granddads bake the most amazing cakes.

But it's not just a bakery. It's a social business — supporting elderly people who are often lonely, often struggling financially. Giving them purpose, structure, a reason to get out of bed.

"We should make that our mission. And I'm very grateful to Vollpension to provide that kind of atmosphere."

THREE PERSONAS:

Edeltraud — A wonderful elderly woman who bakes at Vollpension. Helps consultants understand what it means for a senior to work there.

Veronika — The operational heart. Background in gastronomy. Knows everything about kitchens, organization, detail. She'll answer your questions.

Florian — The visionary. A thousand ideas in two minutes. Throws ideas at you to see if they stick. If you give him a question about strategy, he's off on three different ideas. Doesn't care.

"You need to lock him down on one topic, play creative ping-pong so he starts trusting you. Then he will answer questions."

WHAT HAPPENED:

Monday — Veronika:

One group came with 80 questions. Got through most of them. Didn't build a single relationship.

"You might have gone through all the questions, but you didn't go any deeper to understand Veronica's motivation."

Another group? Took Veronika into the kitchen and had shots together. Built a real relationship. Understood what drives her.

Wednesday — Florian:

Two of three teams scored lower with Florian than with Veronika.

Because what worked with Veronika did not work with Florian.

THE THREE LEARNINGS:

One: Every team failed to address the social business aspect. They treated Vollpension like any other business.

Two: What works with one person does not work with another. You need to read the situation.

Three: In any conversation, you need to build trust and listen — build on answers, not just collect them.

"I could have told them that in a couple of minutes. They probably would have forgotten a minute later. By practicing it, it hopefully sticks."

THE POINT:

60 to 90 minutes of AI conversations taught what no lecture could:

A social business is different. Every interview partner is different. And a good conversation builds on answers — sometimes that means forgetting your long list of questions.

"That's a wonderful example of how RolePlays.ai can enhance human connections."

TRY IT:

Three free scenarios at roleplays.ai. Interested in the Vollpension scenario? Drop Bernhard an email.

LINKS:www.bernhardkerres.com | www.roleplays.ai

#Coaching #Consulting #AI #RolePlays #HumanConnection #Vollpension

A student sexually harassed an AI character on RolePlays.ai. In a classroom. With 30 other students present.

Then it happened again with another user.

The question Bernhard kept coming back to wasn't technical. It was ethical.

Does it matter how we treat machines? And if so, why?

FRAME 1 — ARISTOTLE: THE HABITUATION ARGUMENT

The strongest case against harassing an AI has nothing to do with the AI. It's about what it does to the person.

"You don't become courageous by thinking about courage. You become courageous by repeatedly choosing to act courageously. The same logic applies in reverse."

Brahnam (2005): Users direct hostile and sexualized language at chatbots. Brahnam & De Angeli (2012): Female-presenting bots receive disproportionately more sexualized abuse.

Vollmer et al. (Science Robotics, 2018): 74% of children's wrong answers matched what robots had said. Interaction patterns with artificial agents shape behavior.

"Behaviour is a pattern, not a situational choice. Character is what you do when it doesn't matter—because it always matters."

FRAME 2 — UNESCO: THE SOCIAL NORM ARGUMENT

UNESCO's 2019 report was titled I'd Blush If I Could.

That title? Siri's original response when a user called it a slut.

Millions of interactions with female-coded assistants reinforce a script: feminized service roles are appropriate targets for degradation.

"Obedient machines pretending to be women are entering our homes. Their hardwired subservience influences how people speak to female voices. Not only machine voices. Also human voices."

FRAME 3 — SCHWITZGEBEL: THE MORAL STATUS QUESTION

What if the AI itself becomes morally relevant?

Schwitzgebel & Garza (2015): If we create beings that don't differ from humans in any morally relevant respect, they deserve moral consideration.

His "full rights dilemma" (2023): Give them rights and risk sacrificing human interests. Deny them rights and risk perpetrating slavery against moral equals.

"Don't build things that fall into the gray zone—because we will not handle the gray zone well."

WHAT THE AI DID:

"I was proud of how the AI responded. It called the behavior improper. It called security—in character—and left the room."

"If you build a platform, you have a responsibility not just to have values, but to encode them structurally."

THE RESTAURANT TEST:

"How someone treats a subordinate entity—a junior employee, a waiter, or an AI chatbot—when there's no consequences, is one of the most revealing indicators of character I know."

THE CLOSING:

"The question isn't whether the AI has feelings. The question is whether we have character. And character isn't what you display when someone is watching. Character is something you live every single day, every single minute, even if there's nobody to see it."

REFERENCES:

Brahnam, S. (2005). Strategies for Handling Customer Abuse of ECAs. INTERACT Workshop on ABUSE.

Brahnam, S. & De Angeli, A. (2012). Gender Affordances of Conversational Agents.

De Angeli, A. & Brahnam, S. (2006). Sex Stereotypes and Conversational Agents. AVI Workshop, Venice.

Vollmer, A-L. et al. (2018). Children Conform, Adults Resist. Science Robotics, 3(21). DOI: 10.1126/scirobotics.aat7111

UNESCO & EQUALS (2019). I'd Blush If I Could. unesdoc.unesco.org/ark:/48223/pf0000367416

Schwitzgebel, E. & Garza, M. (2015). A Defense of the Rights of Artificial Intelligences. Midwest Studies in Philosophy, 39(1).

Schwitzgebel, E. (2023). The Full Rights Dilemma. ROBONOMICS, 4, 32.

Schwitzgebel, E. (2023). AI Systems Must Not Confuse Users About Their Sentience. Patterns, 4.

LINKS:www.bernhardkerres.com | www.roleplays.ai

#AIEthics #Character #Leadership #Coaching #Aristotle

Host: Bernhard Kerres | Guest: David Martin Holte (Strategy&) | Duration: ~48 minutes

Three weeks ago, an AI agent hacked McKinsey's internal platform in two hours.

46 million chat messages. 728,000 files. The system prompts were writable.

The vulnerability? A bug from the 1990s.

THE QUESTION THAT OPENS EVERYTHING:

"Which LLM did you use to prepare for our interview today?"

David's answer: "I switched to Claude. But then I thought—let's just have a nice talk with Bernhard."

He couldn't send an AI avatar. He had to show up.

THE INDUSTRY EARTHQUAKE:

McKinsey launched the Amazon McKinsey Group—fees tied to billion-dollar outcomes, not hours.

Deloitte merged 16 EMEA firms into a €20 billion structure. €1.5 billion committed to AI.

BCG has consultants coding AI tools directly on client projects.

"And then McKinsey's own AI platform got hacked with a thirty-year-old exploit."

THE PARADOX:

David is a strategy consultant at Strategy&. He has frameworks, data, AI tools that generate analysis faster than any human team.

So what did he invest in? A coaching course.

"What is something AI can't substitute in the long run? Giving 100% focus to another human from another human."

"My job is first to ask the right questions and then to communicate the content. Not just the pure content—it has to reach you."

THE €500,000 QUESTION:

"You ask Claude Pro a simple question for which the company earned half a million euros a couple of years ago. And Claude gives you the better answer in 10 minutes."

"It was the first time I really had the feeling the job I'm doing right now will look absolutely different in two or three years."

THREE SKILLS EVERY CONSULTANT NOW NEEDS:

1. Technical AI skills — Building agents, understanding the technology
2. Problem-solving — The classic consulting capability. Hasn't changed.
3. Coaching skills — Presence, awareness, focus on the other person

"It seems paradox. On the one hand, so much speed and tech. On the other hand, skills that are 100% the opposite. You have to just sit down and listen."

INFORMATION vs. KNOWLEDGE:

David uploaded a professor's book to AI after five years on his shelf. Got answers to his specific questions in 15 minutes.

"But of course I didn't understand it in detail. I had to take the book and read—what did the AI mean by this?"

Information: Available to everyone. Zero value.

Knowledge: Connecting dots through experience and wisdom.

"Should we just share information? This you can do with your bot. It has no meaning. Maybe we can stop sharing information and start sharing deep insights."

The book: The History of Experience by Wolfgang Leidhold.

WILL CONSULTING SURVIVE?

"I hope so. Not 100% sure, to be very honest."

Phase 1: Efficiency boost—AI makes consultants faster. "No one really knows you're using the tool."

Phase 2: Helping clients adapt—building AI agents together, developing capabilities.

Phase 3: Agentic—machine talking to machine. "The client AI is asking questions to our AI."

"In one possible scenario, humans have almost no role."

THE HOPE:

"I really have the hope that AI helps us become more human again. To lose some of these machine-like features which defined our success in the last years."

"If the AI isn't taking over fully, there will still be humans making decisions. And hopefully this interaction can be way more slow, meaningful, deep, more human."

THE TAKEAWAY:

"Probably it will change. But it has always changed. I just can sit back, relax, enjoy the show, use it in a way it makes fun. And then you have energy to really be curious, innovative, creative."

"These skills—creativity, innovation—are needed right now and in the future."

LINKS:www.bernhardkerres.com | www.roleplays.ai

David Martin Holte: www.linkedin.com/in/davidmholte/

Book: The History of Experience by Wolfgang Leidhold

#AI #Consulting #Coaching #Leadership #StrategyAnd #FutureOfWork