The challenge: knowledge that lived only in people's heads
The event ran on tribal knowledge. How each of a dozen volunteer roles actually worked sat with the people who had done them before, passed on in meetings and repeated from memory every year. Nothing was written down, so every season started by reassembling what the last one already knew, and a single departure could take a role's knowledge with it.
What we built
We turned the planning conversations the team was already having into documentation, without adding a writing job on top.
Record and transcribe
Existing planning meetings were recorded and transcribed, so the raw knowledge was captured where it was already being shared.
Organise with AI
AI sorted the transcripts into role-by-role procedures - what each volunteer does, in what order, and what they need to know.
Human review
The organisers checked and refined the drafts, correcting anything the AI had misread and adding the detail only they knew.
The result
From a few hours of work, the event has a 40-page manual covering twelve volunteer roles - written once, reusable every year. The knowledge now lives in a document the team owns rather than in the memory of whoever happened to do the job last time.
Why it matters
Most organisations carry the same risk: the way the work actually gets done is undocumented, and it walks out of the door when people move on. Capturing it from conversations people are already having is far cheaper than writing it from scratch, which is why it tends to get done at all.