This week's episode is a case study in what AI looks like when it's doing real work.

Juan Garcia runs an insurance company in Spain. Industry average profit margin is 5%. He's at 15%, headed for 18%. The difference? Five AI agents in production doing real work. Not pilot projects. Not demos for the board. Actual agents handling claims, customer questions, marketing decisions, fraud detection, and underwriting. His claims adjusters went from 10 cases a day to 50 because the AI does everything except the stuff that actually needs a human.

Here's the thing. Juan started this in mid-2023 with GPT-3.5. His team built 75 subagents to control quality on that first chatbot. That's the kind of smart engineering that makes AI work in production. He'll tell you exactly when to let the AI decide, when to kick it to a human, and why confidence thresholds matter more than anyone talks about. He'll also tell you where they won't use AI. Rejecting claims. Handling money. Anything that needs actual empathy. You can't fake that and you shouldn't try.

Want to know what works in production? Juan's got the decisions and the profit margins to back it up.

Also in this episode: Juan's Tuio presentation

This week's episode breaks the usual format, and that's the point.

Instead of a guest or a debate, Rob does something he hasn't done publicly in a long time. He reads the foreword to a book he's actively writing. The first one since 2015.

Back then, his books helped define how people learned Power BI. For a few years, he was literally the guy who wrote the book. Then he stopped. No updates. No sequels. An entire generation of practitioners came up without ever encountering his work.

So why return now?

Because the same pattern is repeating itself, just louder. This time with AI. The hype is everywhere, the confusion is real, and business leaders are being handed tools without a usable mental model for how success actually happens.

This foreword is an explainer. Plain English. Business focused. Written for leaders and for the people who have to help those leaders make good decisions. No formulas. No technical flexing. Just a clear frame for thinking about AI in a way that doesn't implode six months later.

Consider this episode an early access audio version of something that's still being built.

Give it a listen. And if the foreword resonates, stay close. This may not be the last chapter you hear early.

Those Excel macros running your business were never meant to be permanent. Someone in accounting built them because the company needed custom software and didn't have the budget or patience for a two-year IT project. IT hates them. You know they're fragile. But they work. And compared to expensive software that never quite fits, working counts for a lot.

In this episode, Rob and Justin dig into what might finally replace that world. Not in theory, but in practice. Over the next four years, is the real shift AI helping people build traditional software faster and cheaper? Or is it software that actually has AI running inside it at runtime? The difference matters if you're deciding where to invest time, money, or political capital.

They also tackle who's going to build this next-generation line of business tools. Is it the Power BI crowd all over again? The VBA veterans reinventing themselves? Or a new kind of builder who sits closer to the business than IT ever could?

If you're nursing a mission-critical spreadsheet you're afraid to touch, or paying too much for SaaS that almost fits, this conversation will feel uncomfortably familiar. And useful.

Listen to the episode and start thinking about what replaces your macros before they replace themselves.

Every week brings a new AI model, a new benchmark, and a new reason to believe everything just changed. But for most companies, none of that matters if the people closest to the work can't use these tools to build something real.

In this episode, Rob and Justin walk through what democratized data science really looks like. Not dashboards. Not prompts. Actual analysis and custom software built around a specific problem, driven by someone who knows the data well enough to challenge the answers. The difference isn't the technology. It's the person driving it. Someone who understands the data, the domain, and how to spot bad answers before they turn into bad decisions.

That's where the data gene shows up again. When those people are empowered to build software fitted to how work happens, off-the-shelf tools stop feeling helpful and start feeling like friction. This episode is about noticing that shift while everyone else is still watching benchmarks.

Be sure to subscribe on your favorite podcast platform for weekly reality checks on AI and Analytics delivered straight to your inbox.

Every week brings a new AI benchmark. Higher scores. Bigger claims. Louder voices insisting this changes everything. And yet, when you put AI in front of a real business problem, none of that noise seems to help. In this episode, Rob and Justin dig into why AI benchmarks often feel strangely meaningless in practice and why that disconnect is the point. Benchmarks aren't useless. They're just answering a different question than the one most businesses are asking.

This isn't just random conjecture either. Rob walks through what he's learned building actual AI workflows and why a twenty percent improvement on a leaderboard rarely translates into anything you can feel on the job. They talk about why model choice usually isn't the bottleneck, why swapping models should be easy if you've built things the right way, and why the most successful AI work rarely shows up as a flashy demo. Most of the value is happening quietly, off-screen, inside systems that look a lot more like normal software than artificial intelligence.

Rob and Justin also talk about why explaining AI is often harder than building it. The first demo people see tends to stick, even when it's the wrong one. Consumer AI feels magical. Business AI face plants unless it's built with intent, structure, and real context. This episode gives leaders better language for that gap, without hype or panic. If you're done chasing benchmarks and just want a way to think about AI that survives contact with reality, this episode's for you.

Everyone keeps asking whether AI kills Power BI or makes it stronger. Rui Romano flips that entire question on its head. As the Microsoft PM behind PBIP, TMDL, and all the file format work that rebuilt Power BI's foundation, he explains how the platform accidentally became one of the most AI-ready systems in analytics - and it wasn't by accident, not really. His team was solving problems for real developers who were tired of unsupported workarounds and offshore relay races. They weren't training agents. But the work they did means AI now feels native instead of duct-taped on.

What we learned was that the semantic model is still the highest ground in this whole space. While other tools let AI stumble through raw tables and pray the math holds up, a proper model gives AI the one thing it absolutely cannot fake: context. Relationships. Business logic that works at every level of granularity without falling apart. Rui breaks down why that matters now more than ever, why all the hardening work his team did keeps your models from exploding when an agent gets ambitious, and why the future of BI isn't about cranking out another hundred pixel-perfect dashboards. It's about fast iteration, lower friction, and answers you can trust at scale. Dashboards still matter - but only the ones people use.

This conversation goes deep on architecture, not hype. Rui talks about what's changing right now, what still needs work, and why natural language will eventually beat drag-and-drop for a lot of what we do today. If you've been wondering whether to invest in real semantic modeling or just let AI figure it out from scratch every single time, this episode makes the case for why foundations always win. Always.

Listen in and get ahead of the shift. And if the episode lands for you, leave us a review to help other folks find the show.

This episode starts with a familiar scene. A role opens, the applications pour in, and suddenly you're staring at a mountain of resumes that deserve real attention but arrive faster than anyone can process. The mix had everything… experienced candidates, newcomers trying to break in, and a growing stack of AI-generated submissions that looked sharp until you asked a second question.

That's where Haystack came in. Instead of using AI as a blunt filter, Rob and the team treated it like a collaborator. Teach it what matters. Teach it what P3 looks for in a teammate. Teach it how to separate real signal from polished noise. What came back wasn't a robot recruiter. It was clarity.

And Haystack is only half the story. As the conversation unfolds, Rob and Justin zoom out into the broader pattern they're seeing across all the small, useful agents taking shape inside P3. The stuff that isn't blind hype. The stuff that quietly fixes overloaded parts of the business and makes the human decisions easier to get right.

Because that's the through-line here. When AI handles the overflow, people get to spend their time on the work that actually requires judgment.

Queue it up and hear what happens when AI stops pretending to be magic and starts doing real work. And if you've got a corner of the business that's begging for that kind of clarity, we can help you find the tiny build that changes everything.

Rob finally cracked his years long standoff with the podcast lair cat, and the fix was hilariously simple. That small victory ends up setting the tone for the whole episode, because everything that follows has the same energy: real problems that only make sense once you shrink the solution down.

As Rob talks through the cat truce, Justin brings in a different kind of chaos. A customer service bot that sounded fully in command yet never actually did the thing it said it did. Pair that with a hiring queue full of AI written applications, and the whole picture starts to come into focus.

Once you see the pattern, you can't unsee it. The wins only show up when the AI job gets small. The fantasy football tool works the moment AI stops trying to scrape the entire internet and instead only writes the human part. The hiring filter works when AI is there to catch repetitive patterns, not run the whole show. Even the experiments coming out of Danielson Labs click only because the AI calls are tiny and the real work sits in regular code.

Everything points in the same direction. Let AI handle the one thing only AI can do, then let normal tech take it from there.

Everyone's talking about AI like it's plug-and-play. Spoiler: it's not. In this episode, Rob digs into why Big Tech's billions in AI R&D haven't yet turned into matching revenue — and what that means for the rest of us. The truth? The real business wins don't come from off-the-shelf models; they come from smart customization. Rob breaks down the "magic Lego brick" approach that separates hype from practical reality, showing how everyday tools like Power BI and Power Automate can connect to AI in surprisingly simple (and powerful) ways. He also revisits Bill Krolicki's "Vendor Bot" example to prove that you don't need to be a researcher or a billionaire to make AI deliver real results. If you've ever opened ChatGPT, asked it to "optimize operations," and gotten nowhere — this one's for you.

When the bubble pops, what's left standing?

Everyone's calling AI a bubble — and yeah, there's some wild stuff happening behind the scenes. Billions changing hands between the same few tech giants. Investments that look more like circular loans than innovation.

But what if the real bubble isn't AI at all?

Rob breaks down what's really inflating, what's not, and why your business shouldn't lose sleep over it. The headlines may sound like the sky's falling, but the tools driving real results aren't going anywhere.

Some bubbles burst. Others just float higher. Listen now to hear Rob's take on which kind this is and why the real value of AI doesn't care what the stock chart says.