The Rest Goes Stale

A team of twelve people knows an enormous amount. Each of them carries years of client meetings, decisions, mistakes and shortcuts. The problem is that none of them knows all of it. And what they know together lives nowhere. It is scattered across calendars, mail threads, notes no one opens again and recordings no one had time to listen to. Most of it they carry in their heads. And the rest goes stale.

When a team like that talks about AI, the conversation is almost always about the model. Which one is smartest, which one writes best, which one shipped last. I have spent hundreds of hours on this and my experience is a different one. The model was rarely the bottleneck. The bottleneck is that the team's own knowledge is never gathered in one place where it can actually be used.

Capturing is the easy part

Capturing information is no longer hard. Record the meeting, transcribe it, drop a note. Anyone can do it today and most tools do it for you. That is exactly why that part is not worth much. Everyone has it.

What is hard, and what is actually worth something, is what happens next. That what you captured gets structured and connected into a whole you can use again and again. That you can ask a new question of the same meeting tomorrow and get a different answer, because your question is a different one by then. That a thought you had in the car does not become an isolated note but is woven into everything you already know.

A pile of recordings is not knowledge. It is raw material. Knowledge is when the material hangs together.

What a team loses

In a team, that scatter costs more than you think. The person who sat in that client meeting two years ago remembers the risks. When she is on holiday, or has left, those risks are gone. The next meeting with the same client starts from zero. Not because no one cared, but because what no one wrote down in the right place never became part of what the team remembers.

That is what quietly aches, every week. No disaster. Just the same questions asked again, the same lessons learned again, the same context built from scratch every time. A team that produces more but remembers less.

A system that understands how everything connects turns that around. The client meeting already knows which similar projects you have run and what went wrong last time. The recording links to the discussion you had six months ago. Knowledge stops living in individual heads and starts belonging to the team. It gets smarter every time you capture something, instead of fading.

Proof, not a claim

This is not an idea I think sounds nice. I have built it and I run it every day. A system that takes my conversations, meetings and notes and turns them into connected knowledge I can question, instead of a pile of files I forget. It is called deep-thought. Three years of building sit behind it, in production, in Sweden.

What I have learned from running it that long is simple. The hard part was never getting hold of a good enough model. The hard part was getting what I already know to stop running through my fingers.

What you already know

A team rarely needs to know more. It needs to reach what it already knows, gathered, connected and alive, instead of scattered and on its way to going stale.

The model gets smarter every month. It matters less than you think. The question that decides things is whether what you learned today still means anything tomorrow.

See also: Each Meeting Alone Drives Nothing (series 19), What if we'd recorded this? (series 36) and Text Is Data (series 8).