Turning a messy stakeholder conversation into a clean decision memo.

You know the conversation. A room of eight people. Three priorities that don't quite line up. Two strong personalities pulling in different directions. A budget question nobody wants to be the first to ask.

Forty minutes of discussion, ten minutes of agreement, and a vague sense that you'll follow up over email.

You leave with a notebook full of half-sentences, an inbox full of incoming context, and a sinking awareness that if you don't capture the decision now, the room's memory will splinter into eight different versions of it by next week.

The traditional move is to stare at your notes until they organize themselves. They never do. The new move is faster and produces a better result.

The workflow

When I come out of a meeting like this, here's what I do.

I dump everything into an AI tool. Raw notes. Half-thoughts. Names with arrows. Quotes I jotted down. I don't clean it up. I don't reformat. I don't try to make sense of it first. That's the AI's job.

Then I ask for a decision memo with five sections.

1. The decision we made (or, if we didn't make one, the decision we still need to make)
2. The reasoning behind it
3. The next steps and who owns each
4. The points that are still unresolved
5. The dissenting views worth keeping on the record

That's the prompt. Five sections, one page, plain language.

The first draft is usually 80 percent right. The wrong parts are usually in section 1. The AI doesn't always know which decision was the real one versus the side conversations. I correct that, ask it to regenerate the rest with the corrected anchor, and within ten minutes I have a memo I can send to the room.

What this does that the old way didn't

The old way of writing meeting notes had a hidden cost. It privileged whoever had the cleanest handwriting and the freshest memory, usually the person closest to the decision, who already knew what they wanted the memo to say. The notes became the decision, not the other way around.

The AI-assisted version surfaces what was actually said. It isn't perfect (it'll miss subtext and tone), but it captures the substance more evenly than any single person's notes would. The dissenting views section in particular is the one I'd skip if I were writing the memo by hand. The AI doesn't have a stake in being right, so it surfaces them.

The other thing the AI version does is force structure. The "still unresolved" section is the one most leaders quietly delete from their hand-written memos because admitting open questions feels like admitting failure. It isn't. Naming what's still open is the fastest way to actually close it.

The leadership note

The temptation, the first few times you do this, is to let the AI write the whole memo and send it as-is. Don't. The AI organizes the thinking. You make the decision.

There's a real difference between here's what the room said and here's what we're doing about it. The first is a transcript. The second is a leadership document. The AI is good at the first and adequate at the second. Your job is to add the layer that turns the transcript into the document: the judgment about which trade-off was acceptable, which voice in the room carried the most weight, which next step is non-negotiable.

That layer is what your team is paying you for. AI just gets you to it faster.

What "good" looks like

A clean decision memo is short. One page. Five sections. Written in language a person who wasn't in the room can understand. Sent within 24 hours of the meeting, not next week. Includes the dissenting views, not because you have to, but because the people who held them will respect the memo more if you do.

If you're getting decision memos out faster, with cleaner sections, that include the open questions and the dissent, your meetings are getting more productive even if nothing about the meeting itself changed.

That's what AI workflows are actually for. Not replacing the meeting. Making the meeting count.