🎙️ 0 — Meet
Meet captures client needs from conversations, interviews, and workshops. Keep the original words and their source so every later stage can check the client's intent.
Purpose
Capture client intent without losing it through hand-offs.
| Input | Output | Owner |
|---|---|---|
| Client dialogue, stakeholder discovery sessions, voice and video calls, incident debriefs | Transcripts, meeting summaries, highlighted domain quotes, and customer intent anchors | Client + Product Owner |
What is lost without it
In traditional delivery, a client's words pass through several people before a Developer sees them, and each hand-off keeps less.
| Traditional ticket | Meeting-Driven backlog issue | |
|---|---|---|
| What the client said | "Our regional managers spend 45 minutes every morning cross-referencing CSV exports from SAP with local inventory sheets before trucks can roll." | The same sentence, quoted verbatim, with a link to the exact moment in the transcript |
| What the Developer receives | "Improve inventory export" | A named problem, who is blocked, and what it costs them |
| What is gone | The 45 minutes, the SAP source, the trucks waiting, the people affected | Nothing |
| How it can be checked later | It cannot. The only record is someone's summary | Every later stage can be checked against the original words |
Tests cannot recover missing business context. Preserve the client's words before summarizing them.
Core capabilities
1. Transcription & speaker attribution
AI transcribes meetings with timestamps and speaker names.
2. Verbatim intent anchoring
Extract exact quotes and link them to domain concepts. Triage cites those quotes as evidence:
"Our regional managers spend 45 minutes every morning cross-referencing CSV exports from SAP with local inventory sheets before trucks can roll." — Operations Director, Meeting Transcript (2026-09-24, 15:14:32)
The Meeting Pack artifact
At the conclusion of each meeting, the intelligence pipeline produces an immutable Meeting Pack:
- Full transcript: Verbatim text indexed by timestamp and speaker.
- Executive summary: 1-page summary of business drivers, urgency, and core challenges discussed.
- Domain concepts & terminology: New terms or acronyms introduced by the client, staged for commit to Persistent Context.
- Candidate business needs: Needs ready to be qualified into requirements and ranked as backlog issues in 1 — Triage.
Where AI is used
AI captures and structures the dialogue; the Product Owner steers the conversation and the client confirms what was said.
| Task | AI role | Human role |
|---|---|---|
| Transcription | Convert the audio or video stream to a timestamped, speaker-attributed transcript | Verify speaker accuracy |
| Quote anchoring | Extract verbatim quotes and link them to domain concepts | Confirm the quotes reflect the client's meaning |
| Meeting summary | Produce an executive summary, new terminology, and candidate business needs, using the glossary and instructions in Persistent Context | Confirm the problem statements with the client |
Agents and skills
Persistent Context supplies the glossary and Meeting Pack format. New client terms go back into the repository.
Meet runs on the primitives published by coding-pal:
- A Meeting Analyst agent turns the sanctioned tool's plain-text transcript into the Meeting Pack, and prepares the Product Owner's read-back list.
- A meeting-pack instruction keeps the transcript immutable and every quote verbatim and client-attributed.
- A meeting-pack skill normalizes the transcript, validates the pack against the transcript and domain glossary (digest, verbatim quotes, glossary terms, and named human confirmations), and emits one backlog issue per confirmed need for 1 — Triage.
- The domain glossary (
docs/glossary.md) is what the agent reads, and the new client terms it finds are staged for a pull request.
See the coding-pal Meet guide.
Exit gate
A meeting successfully closes when:
- The meeting transcript has been captured, verified for speaker accuracy, and committed to the repository.
- The client has confirmed that the problem statements in the meeting summary accurately reflect their business pain.
- Candidate business needs have been emitted to the backlog for ranking in 1 — Triage.