A Journey Into AI Fluency, with Seventy-Five People

Over the past few weeks, several AI-Team programs wrapped up at some of my client companies. A journey of a few months of 'hiring AI', to understand how it transforms us and what results existing teams can actually deliver. A journey made by organizations far apart from each other in sector, size, and company culture. Which I'm telling you about, for the first time, today!

AI Team

I worked personally with about seventy-five people who, over three months, more or less, did the thing that almost never happens at work: they were given permission to spend time daring, trying, failing.

Teams sent out to scout, experiment, write policy, hunt for data, do high-level FAFO. And to produce tangible things. Even people who called themselves "non-technical" walked away with mini-apps, agents, artifacts. Real objects, made by their own hands and by those of an artificial colleague.

In practice, we met for a couple of days of workshops, and then, week after week, online, we explored a few corners of what it actually means to 'hire an AI' at work. Beyond learning, each team member had the chance to develop their own project: it could be software, an agent, a policy analysis, a dataset... anything, as long as it was built with AI.

I asked companies for one thing, clearly: build heterogeneous groups. Very heterogeneous. The skeptic sitting next to the enthusiast. The twenty-five-year-old next to the fifty-year-old. The expert next to the person who introduces themselves with an apologetic smile, saying "I'm the tech-illiterate one here." I asked for it because I knew what would happen, but watching this 'diversity' play out on such a common subject was, for me, deeply moving.

What I take away from it, in the end, isn't the list of projects. It's a human experience I find hard to describe without sounding rhetorical. But I'll try anyway, and share some of what I observed.

The Relationship With AI: From Distrust to "I Can't Go Back"

At the start there's fatigue, and sometimes anger. Buying into the idea of "having an AI intern do things for you" costs something. You start out wanting to control every word, and you end up delegating an intention instead: it's a shift in posture, not in software, and at first many experience it as a loss.

Some wrote about feeling, in the first few weeks, a real distance between themselves and the tool. Some were wary of "this abstract entity that promises to do everything, maybe even better than a human." And some started out openly critical, not of the technology itself, but of what they feared it represented: one person told me about a son who loves to write, and teenage grandchildren who ask AI about everything without ever straining their memory anymore. Their question, written in black and white, was: "will there even still be a need for any of this, tomorrow?"

Then, at some point, the click happens. Some felt it on day one. For many, it came the day they saw their first agent actually work, and one person wrote that the moment "turned everything around" for them. It's the instant when AI stops being a slightly smarter search engine and becomes a partner you can actually DO things with. One of them put it better than any definition of mine could: "I stopped asking myself what to write in the prompt and started asking myself what I wanted to happen next." That shift says it all.

After the click comes the enthusiasm, and this is where the stories get strange, not always in a good way. The person who had spent a lifetime telling themselves "I don't get along with computers," and now finds themselves building software without ever having learned Python or JavaScript. "I have superpowers now," they wrote, and it's not hyperbole: it's the exact description of what they feel. (More on keeping that under control later: this isn't an initiation into reckless vibe coding.) Others openly say they can no longer go back to how they used to work. Some go overboard, staying up at night because they have too many ideas to build and it takes more than a few sleepless nights to work through them. They probably need help pacing themselves.

Then, third act, sooner or later everyone comes back down to earth. The enthusiasm gives way to a more grown-up question: "Where's the real value here? For me, but also for my team and for the company." "In three months I mostly learned the limits, and that's exactly why I'm now better at using it." Whoever gets here is no longer the skeptic from the start, nor the mid-journey enthusiast. They've learned the hardest thing of all: when to trust it and when to check it. And some hold both things together all the way through, the wonder and the concern, and by the end of the program they're more capable than before but haven't stopped asking themselves the uncomfortable questions about the future, and they understand when not to use AI. That one, to me, is the person who understood it best of all.

It's a world of incredible nuance, impossible to fully capture here in text alone. But I can tell you that the emotion I saw in their eyes during the program, and especially in the leadership they presented their results to, was something unique.

It's Not a Course

Because here's the point that risks getting lost: they didn't take a course. They spent time inside a space where they learned by doing, and, experimenting, brought home small, real projects: on their own, with no consultant standing behind them guiding their hand, at most with help from a slightly nerdier colleague. And so, by doing, they saw with their own hands how work changes. As written above: "it's clear we can't go back anymore."

When AI Shifts the Boundaries Between Roles

I already know what someone out there is thinking: fine, but who actually manages the software once it's in the company? And you have to see what happens once the program ends, in everyday life. Some will get lost, the company won't respond, some will refuse to keep going. That's fine. This isn't the definitive path, just the beginning of a kind of work that needs time to settle. I say it often: AI is extremely fast, but we're slow to absorb it.

What I really want to talk about is the thought that's been on my mind for months, what I call the Full Stack Human, the single person whose range of action widens, I've already written here, and I'd suggest taking a look if you haven't already. What I saw over these three months is the next step: generative AI, at work, doesn't add one more tool to the toolbox. It lets people do new things in new ways, but above all, it shifts boundaries.

Personal boundaries, first of all: every time someone discovers they can do something they've never done before, that person's boundary moves. It stretches their activities beyond the little garden they've always lived in. And, without much worry, you start seeing lawyers dealing with algorithms and economists thinking about user experience, or software developers writing sales playbooks.

And it's clear that working as a group is still needed, to keep things on solid ground, to validate ideas, to accept other 'garden keepers' encroaching on your patch, and to discover that together you can reinvent how you tend the garden itself 😊.

You realize there are alternative solutions out there, valid, effective, and often ones that redistribute new kinds of advantage.

You realize you can produce new kinds of objects, not just spreadsheets to run processes.

AI Objects: What Can You Produce Beyond Excel?

Chaos, you say? A bit, yes. Chaos that needs to be observed, reasoned through, sifted, and eventually secured. Because otherwise you end up back at "that's not how we do things," at "see, it doesn't work?" or, worse, "we've never done it that way for good reason."

Experimenting with AI lets you carry out complex tasks in very little time: why not do it when the benefits are obvious? The advantage of doing it inside structured companies, compared to a solo operator or someone with no one around handling infrastructure and security, is that an imperfect, experimental, yet functioning object can find a home in company systems, as long as whoever is responsible for them recognizes it as a new kind of digital object. I call them AI Objects. Seen from the outside they look the same as the old ones, but they were built in entirely new ways and have completely different lifecycles.

I'm talking about AI Skills, Artifacts, or Vibe-Apps, but also AI Agents (simple or more complex), disposable software, MCP connectors, clean datasets, MD instruction files, and so on.

I know someone out there is screaming inside, "Max, how can someone who isn't technical possibly be aware of what software even means! If you don't know what an SDLC is, how can you build anything durable? Vibe coding is nonsense!"

But it's precisely when a team is heterogeneous, when one function starts doing a piece of work well that used to belong to another, that the results start to get interesting.

If your users become able to create objects, even ones that are architecturally imperfect, and your organization has the capacity and the patience to analyze them, review them, maybe rebuild them, and use them as a clear specification of what the user actually needs, you'll discover new ways to generate value.

And if people use agentic systems to explore their own information systems, and start chatting with the ERP or the CRM to understand or to get things done, and, instead of Excel and PowerPoint, start producing AI Objects, context files for agents, loops... well, you'll also discover how boundaries are bound to change. And not just for the individual or a department, but for the whole company.

It needs to be absorbed.

And here I have to say something uncomfortable. For thirty years we told ourselves that software was designed around people. Almost always, it went the other way around: it was people who bent themselves around the limits of the software. It's because of badly designed software (or software never torn down and rebuilt once its limits were obvious) that we invented processes tailor-made to those limits, often without even realizing it.

Excusatio non petita, accusatio manifesta. I know. I've been guilty of it myself for thirty years 😊

What excites me about generative AI isn't that it lets you automate those processes. It's that, for the first time, it gives you the chance to get rid of them. To redesign work around the person doing it, while keeping the rigor and standardization that make processes solid. For the first time, we can build ten versions of the same interface in the time it used to take to build one, and then figure out which one works best. We can finally automate processes in places other than Excel while keeping data secure and consistent. Automating a process that was born crooked doesn't get you far. Removing it and replacing it with different dynamics can create real value.

But you need to absorb the new way of working, define boundaries, extract value.

So, Now What?

And here's where I get to the point, the one I wrote everything else for.

All of this is a journey. A journey into your own AI fluency. And journeys, by definition, can only be taken, only partly shared (no matter how many photos, videos, or personal stories you tell).

The conclusion is simple, and the seventy-five stories prove it line by line: if you don't raise your own AI fluency, you can't really understand what AI is for.

Whoever uses it "just to summarize texts," without ever getting their hands dirty delegating part of their actual work to it, will never see its potential. If you don't understand what you're dealing with, you can't change how you work, personally, as a team, or as a company, simply because you don't know what to change or how.

And so here's the last link, the one few people want to hear: no consultant can ever tell you, from the outside, what to do with AI in someone else's job. The person who does that job has to figure it out themselves. On the ground. And only then can they get help getting there.

I say this precisely because it's my own job, the job of telling these stories, that's being called into question. The value of AI doesn't transfer the way information transfers on a slide. It's crossed the way an experience is crossed. What a serious program can do is start the engine, give you a method, keep the rhythm, and stop you from getting lost. Only the person driving finds the destination.

The participants' criticisms of AI Team, seen through this lens, say exactly the same thing. Almost all of them ask for one single thing: more time, more practice, more cases drawn from real work. No one asked for more content, more theory, more slides. They asked for more journey. It's the smartest request anyone could have made me.

By the end of the program, on a scale where only nine and ten count as a passing grade, the average came in close to nine. That's a good number. But it's not the number I take home. I take home the skeptic who became the most rigorous and AI-literate person in the group, the enthusiast who learned to doubt, the frightened person who now holds both wonder and the right question together. I take home seventy-five boundaries that shifted.

That's what the journey into fluency is. You can't buy it, you can't delegate it. You have to do it.

When the program ended, a manager wrote a line to their team that stuck with me. Not "continuous improvement," they said, but something more: it's about breaking with the past, with "we've always done it this way." It's about "breaking the rules."

And this is exactly the point where you realize you've actually hired AI. Not when you signed the licenses, but when you started seeing the outcomes of the work of a new kind of colleague: one that changes the work of everyone around it.

It's not a matter of enthusiasm or fear. It's that, once you understand certain things, going back simply stops making sense: processes have already changed, old problems now have more than one solution, projects get thought through differently. What's left to do now is on us: to stay open, and to find a way to make it truly work.

Massimiliano