Every team has years of knowledge they’ve never written down. Some of it is obvious stuff, buried in old projects and folders nobody opens. But the valuable part isn't the facts. It's the judgment. It’s how the team decides, the elements they choose to ignore and what "good" looks like to them. A lot of that has never been said out loud because it’s so deep in their knowledge it’s just how they work.

​

The first move is to get it out of their heads with a working session. Sit everyone down, ask the right questions, write up what they say. But that can’t be the only step because people can't tell you the rules they don't know they follow.

​

There's another source, and they’ve already shown you all of it, in the work they’ve made.

​

We did this for an engineering department. In building a system for them, a core piece we needed was their identity. The values they stand for, how they talk about their work and what’s unique to them versus everyone else in their field. Typically, there’d be an identity workshop and, over several sessions, we’d have arrived at some key values that may have sounded accurate but always carried a bit of, “Is this really true or does it just sound nice?” A better and faster way was to extract it out of their proposals, their site and other internal pieces. All the identity information we needed was already there in what they wrote and what they didn’t write. It was consistent. It was authentic. And with AI, it was extractable.

​

Extracting data is the easy part of structuring a team's thinking. The harder, more important part is how a team makes their decisions, and that can't be pulled from any single document. A conversation helps, but the proof is in the patterns of the extracted data. Look at a team's work against the challenge they were solving and who the client was, and unwritten rules start to surface. For the engineering team, proposals for existing clients didn't lead with their past work together. Just a humble nod to it, with the focus kept on the current ask. A new client, though, got the full résumé. It wasn’t written down anywhere, but they followed it every time. One pattern is a coincidence. But a connection of patterns is thinking.

​

Conversations with the team should still be had, and were had with the engineering team. But they take a different shape. Going into a conversation with a broad, “how do you think?” is much harder to answer than going in with patterns for them to talk about.

​

The result of all of this is more than a database. It's the team's judgment, pulled out of where it was hiding and reworked into something a system can actually use. And that's the difference between AI that gives you a generic version of the work and AI that’s a working part of the team. It’s more than a tool or a better model. It’s the team's own thinking and it’s all right there in the work they’ve already done, just waiting for someone to connect it.