What it looks like
We start by reading the lot together: PRDs, requirements, call transcripts, help docs and how the code actually works today. AI reads it all at once, so our first conversation can be about the gaps instead of catching up.
In the work
VETDSS. PRDs, requirements and a TAFE discussion guide analysed together, plus a field-level map of the data across three systems: which one owns each field, and which way it syncs.
AI RPL. A readiness gate before any build. Is the problem defined without the word AI? Can assessors describe good? Would 80% accuracy beat today? Do we hold real examples? It’s now a skill, so every AI feature gets the same questions.
Coming back to a project. Keeping one memory file per initiative means coming back after weeks away takes minutes, instead of half a day digging through threads.
What comes out
What we know, and what we don’t.