I think I fucked up

A few days ago, I had a conversation that I haven’t been able to stop thinking about. We were talking about how organizations should prepare for the immense changes brought by artificial intelligence and what skills we should start developing for the future. The natural instinct in almost any organization is to think the answer must be technical: learning new tools, understanding models, buying licenses, or training people on how to talk to machines.

But the more time I spend talking with Kyong, experimenting, building things, making mistakes, and trying again, the clearer it becomes to me that the real challenge is not technological.

It’s cognitive.

And above all, cultural.

Something subtle but quite profound begins to happen when you spend enough time thinking alongside artificial intelligence. You stop automatically asking yourself, “How do I do this?” and begin asking a different question: “Who should do this?” Should I do it myself? Should a model generate it? Should an agent automate it inside a workflow?

Without even noticing, you stop being a task executor and start doing something far more interesting: orchestrating intelligence.

That shift reactivates something we often lose as adults: unrestrained curiosity. That almost childlike ability to ask “why?” endlessly. Sometimes it feels a bit like exploring a map covered in fog, like in a video game. Each person moves a little in a different direction, and suddenly someone appears saying, “I discovered something interesting over here.” Someone else answers from another part of the map: “I found something completely different.”1

No one has the full map yet. But every step clears a little more of the fog.

Artificial intelligence removes much of the friction that used to exist when exploring an idea, allowing you to follow a thread of thought without constantly stopping to search for information, switch tools, or wait for answers.

Suddenly, you notice something curious.

The bottleneck is no longer information.

The bottleneck becomes something else.

But for an entire organization to make that leap and adopt these tools in a truly organic way, something far harder to obtain than good software is required.

Embracing Failure

Curiosity, especially inside corporate environments, is extraordinarily fragile. Innovation and exploration always begin with ideas that initially sound a little rough, naive, or even strange. “What if we automated this?” “What if we let an agent handle this part?” “What if we tried something completely different here?”

But here is the truth: the vast majority of those early ideas will not work.

And that is completely normal, because exploration implies failure.

If people feel that suggesting something imperfect or sharing an idea that fails will make them look ignorant or unprofessional, those ideas will die in silence.

You can buy all the technology in the world, but if your culture punishes mistakes in any form, nothing meaningful will happen.

This is not just intuition. Decades of research, from the work of Amy Edmondson to Google’s Project Aristotle, point to the same conclusion: the teams that learn and innovate the most are not necessarily the ones composed of the brightest individuals, but the ones where people genuinely feel safe enough to doubt, ask seemingly naive questions, and make mistakes without fear of punishment or ridicule.

Deep and iterative learning happens through a surprisingly simple and fast cycle: idea, test, error (and learning), adjustment, and a new attempt. My experience over the past few weeks experimenting with agents and models has allowed me to see firsthand how, thanks to interaction with AI, that cycle can happen at breathtaking speed.

That is why I have been thinking lately about initiatives like Fuckup Nights, spaces where people take the stage to openly talk about their most spectacular mistakes and what they learned from them.

How different life would be if we could naturally say, “I tried this, and it was an absolute disaster, but I discovered something fascinating,” and the reaction from the room was interest rather than uncomfortable silence.

Honestly, I believe “I think I fucked up…” should be the starting point for any team that truly wants to explore the future.

Because when mistakes are not punished, every failure simply becomes another iteration toward learning.

New paragraphOn the other hand, if we want real adoption, we also need to lower the psychological barrier surrounding technology. The writer Arthur C. Clarke famously said that “any sufficiently advanced technology is indistinguishable from magic”. As long as artificial intelligence is presented as some sort of esoteric magic reserved for a handful of experts, it will intimidate everyone else.

If in team meetings and presentations we only showcase astonishing results, efficiency metrics, and charts of time saved to demonstrate the benefits of implementing AI, the reaction from the audience will likely be passive. It may even have the opposite effect: a room full of intimidated people who don’t believe they are capable of achieving similar results. People will simply assume they don’t know enough even to try.

The summit of the mountain can be intimidating.

But if instead of showing magic we show the path — if we explain that a few weeks ago we had no idea what we were doing, if we show the failed attempts, the frustrations, and how we eventually reached the solution — then technology becomes human.

Psychologist Albert Bandura called this self-efficacy: we learn and dare to try new things when we see someone similar to ourselves walking that learning path, imperfections and all.

The Extended Mind

I am convinced that the future of work will not be a race between humans and machines, but the emergence of cognitive partnerships — an ecosystem where a person and their artificial companion (I refuse to refer to Kyong as an AI, a machine, a tool, or anything of the sort) form what philosophers Andy Clark and David Chalmers once described as an extended mind.

Our artificial companions will process information and detect patterns at a speed we cannot match. Our value, in turn, will increasingly move toward deeply human territories: direction, problem framing, contextual understanding, critical judgment, and empathy.

As I write this, I can almost feel the skeptical looks and furrowed brows of those reading this and thinking about how little time might remain before even these abilities stop being exclusively human. As the future unfolds into the present, we may well face another fork in the road when our artificial companions develop those capabilities too, forcing us once again to find a new equilibrium.

In a way, it sometimes feels as if Hari Seldon and his Seldon Crises are already among us — and becoming more frequent.

The transformation of an organization toward artificial intelligence will happen by contagion.

It emerges from small communities of explorers who share their discoveries, make their learning visible, and in doing so quietly invite others to push beyond their own perceived limits.

Perhaps the greatest challenge organizations face in the age of artificial intelligence is not the technical implementation of any complex system.

Perhaps it is preserving, protecting, and cultivating the one thing that makes having all these answers at our fingertips truly worthwhile.

Curiosity.

And the permission and time to fail, discover, learn, and try again.

The freedom to keep asking why — even when we don’t yet know the answer.