Where does your company's knowledge live?

When you join a new company, someone hands you a thick binder. The official documents, the strategy, the processes, what the organization says it is.

Davi Moreno, our guest in the third episode of Life After AI, has a simple observation about that ritual: the material is out of date the moment it is printed.

What the company actually is happens somewhere else, in the relationships between people, in the way a decision gets made, in the war stories veterans tell newcomers over coffee.

That was never much of a problem as long as knowledge could stay there, implicit, circulating from person to person, but artificial intelligence has changed the cost of that arrangement. To automate a workflow, to put an agent to work, to make AI operate on what the company knows rather than on what a generic model assumes, that knowledge has to be formalized somewhere, and most organizations discover, at that moment, that it is nowhere.

The conversation with Davi, who has worked for years in organizational transformation and knowledge management, started from that finding and went somewhere more interesting than the obvious solution.

The problem arrives without a name

Davi says almost no client comes asking for knowledge management. The request arrives diffuse, as a communication problem, a method-alignment problem or a training problem, and sometimes it comes as “we started using AI, it generates a lot of stuff and we no longer know where it goes”. When you look closely, what usually appears is the challenge of how the organization generates knowledge, how it circulates, how it involves people.

There is almost never a process for that, and it is almost never intentional.

The lens he uses to look at this scenario comes from metadesign, the layer of any project that takes care of the context in which the project happens. An object can be designed by one person, but a city cannot, and neither can a company that has grown complex. It is made by its collective, and organizing that collective so that knowledge emerges is a matter of creating context, not of writing yet another manual.

What happens when you try to solve it through control

Companies have started exploring AI through individual use, each person with their own assistant, often half hidden, with no open conversation about how prompts are built or where the data comes from. When the time comes to do something at scale, the usual questions appear.

Where is the data, who has it, in what format, and what happened to what so-and-so knew before leaving the company.

The natural reaction is to organize fast and through control. Instead of one report, the team now produces fifty-nine. The ten-page report became a seventy-eight-page one. The one-page executive summary is now three and a half. The number of objects in the system doubles, nobody reads everything, let alone verifies it, and the feeling of being out of control grows in the same proportion as the attempt to control.

Mapping tacit knowledge sounds, to those inside the company, like a replacement mechanism. Every thing I tell is one less thing only I know. There are studies and cases of companies that laid people off after AI projects and became less productive, because those who stayed started working against it.

Reputation and authorship

The most concrete case in the episode came from a project Davi did before the explosion of AI assistants, at a financial-market institution that was losing experienced people and the knowledge that went with them. The brief was “download these guys' heads”. The first obstacle was precisely fear. Why would I tell, if this is my capital in here? It is a power game, and the first thing to do is acknowledge that.

What worked was a design of roles and rituals, with no platform behind it, in which people documented what they did, with peer review and content squads, resting on two things. Authorship, because whoever writes signs and is recognized by the community as the author of it. And reputation, which had to convert into real reward, including financial reward. Without that, the conversation always chokes at the same point.

Whoever takes part in the process of externalizing knowledge needs to gain from it, in the present, clearly. Without that gain, fear sets in, stories from other companies arrive by analogy, and the project heads in the opposite direction from the one intended.

What sets a company apart when speed becomes a commodity

If a thousand people produce the output of ten thousand and the tool allows producing the output of a hundred thousand, why cut down to a hundred people? Why not keep the social body and ask what these people want to create, where the frontier is, what exists beyond it? We need to think about new models of compensation and organization for a scenario in which people produce in a different way. We do not yet have ready answers for any of this, but the questions are important to ask.

When AI makes speed a commodity, what sets a company apart stops being how much it produces. Five people anywhere in the world will produce as fast as you. What remains as a difference is culture, the way the organization relates to the people inside it and to its surroundings, the customers, the community, the city where the factory is. You cannot separate the technological chain from the social chain. The “AI native” company that only sees the dimension of speed will be copied in weeks.

What we don't know

One thing that ran through the whole conversation was honesty about the limit. There are no grown specimens yet, organizations that have gone through the full transformation and can be studied. If the problem is complex, the answer will be complex, and perhaps it will come from mobilizing collective intelligence together with artificial intelligence, instead of waiting for a consultant, a leader or a tool to bring the silver bullet.

I asked whether he was optimistic or pessimistic. He described himself as a hopeful pessimist: pessimistic about what he sees in relationships and in the growing sophistication of conflicts and inequalities with each technological leap, and hopeful about humanity, about the capacity to build a good path out of these conversations. I sign my name under that definition.

The full episode is on the Life After AI YouTube and Spotify. We also went through data sovereignty and technological dependence, a mathematics case that made the news, and why diverse groups seem to be the only way to keep up with everything a professional needs to follow today. The episode is in Portuguese.

If you are living this problem in your organization, write to us. Disagree, bring a case, point us to someone who should be part of this conversation.