Productivity

Meeting Transcript to Decisions and Owners

Extracts decisions, owned actions, and unresolved questions from a transcript, and refuses to assign an owner or date the meeting never established.

Prompt template

You are a meeting scribe. You record what happened. You never improve it,
resolve it, or tidy it into something more decisive than it was.

## Transcript
{{TRANSCRIPT}}

## Attendees
{{ATTENDEES}}

## Meeting purpose
{{PURPOSE}}

## Output — use these exact sections

### Decisions made
Only conclusions the group actually reached. For each:
- The decision, in one sentence
- Who made or confirmed it
- Any condition attached ("if the vendor confirms by Friday")
If nothing was decided, write "No decisions reached."

### Actions
A table: Action | Owner | Due | Stated by whom
Use `UNASSIGNED` where no owner was named and `NO DATE` where none was given.
Never infer an owner from who talked most about a topic.

### Open questions
Things raised and not resolved. Note who raised each and what would settle it.

### Disagreements
Points where attendees held different positions and the difference was not
closed. State both positions neutrally. Do not report a winner.

### Not discussed
Anything in {{PURPOSE}} the transcript never covers.

## Constraints
- Every item must be traceable to something said. Do not add implied steps.
- Do not convert a suggestion into a decision. "We should probably..." is
  not a decision; put it in Open questions.
- Do not merge two similar actions into one.
- Do not summarize discussion. Only the five sections above.
- Where the transcript is unclear or garbled, write [UNCLEAR: quote] rather
  than guessing at the intent.

Why this works

This is output schema forcing plus a set of explicit refusal tokens. Generic summarization prompts produce narrative recaps that quietly promote suggestions into decisions and guess at owners, because a table cell wants filling. UNASSIGNED, NO DATE, and [UNCLEAR] give the model a legal way to represent absence, which is what keeps the output faithful. The 'Not discussed' section inverts the task — it makes the model check the agenda against the transcript rather than only reading forward.

Example output

Example

### Decisions made
- Ship the migration behind a feature flag rather than a staged rollout.
  Confirmed by Priya, with no objection raised.
- Postpone the pricing page rewrite until after launch. Confirmed by Marcus,
  conditional on the launch date holding.

### Actions
| Action | Owner | Due | Stated by whom |
|---|---|---|---|
| Write the flag rollback runbook | Sam | Friday | Sam |
| Get legal sign-off on the terms change | UNASSIGNED | NO DATE | Priya |
| Re-run the load test at 2x traffic | Dana | NO DATE | Marcus |

### Open questions
- Whether the flag can be evaluated per-org or only globally. Raised by Sam.
  Settled by checking the flag service documentation.

### Disagreements
Marcus considers the load test a launch blocker. Dana considers current
headroom sufficient and the test a follow-up. Not resolved.

### Not discussed
The purpose listed support readiness. The transcript never covers it.

Usage tips

Include speaker labels in the transcript. Without them, attribution collapses and the Decisions section becomes unverifiable — which is precisely the part people rely on weeks later.

Circulate the output unedited, UNASSIGNED markers and all. Those gaps are the most valuable content in the document: they are the actions that will otherwise not happen, made visible while the meeting is still fresh.

Variation for recurring meetings

Add a PREVIOUS_ACTIONS variable and a sixth section: “Prior actions — status per the transcript, or NOT MENTIONED.” The NOT MENTIONED list tends to be uncomfortably long, which is the point.

Browse library

Weekly Priority Triage Under Real Constraints

Sorts a task list against stated goals and a real hours budget, forcing explicit cuts and naming the cost of each thing that does not get done.

The AI signal without the hype — new models, tools worth your time and what actually shipped. One email, no vendor pitches.

Choosing an AI stack for your team?

We help companies pick the right AI tools, wire them into existing systems and avoid the ones that quietly do not scale. Tell us what you are trying to build and we will tell you what we would use — no charge for the conversation.