Granola and Claude fix the meeting nobody remembers

You know that moment when a big kickoff call ends and you realize you have no idea who agreed to what? I’ve lived that more times than I’d like to admit. Everyone nods, everyone hangs up, and a week later three people have three different versions of the same meeting.

That’s why this post from a LinkedIn creator stopped my scrolling. The author, who has built and sold a company, shared how the last campaign kickoff they ran happened without a single notebook open. Instead, they let Granola listen to the call and then asked Claude to write the follow-up. The whole recap went out minutes after the call ended, and nobody built the wrong thing.

The problem the author kept running into

Picture the setup. A dozen people on a call: paid media, lifecycle, the website team, a couple of regional teams. Forty minutes in, a date has moved, a deliverable has a new owner, and there’s a fresh deadline for assets. Everyone nods. As the original poster puts it, “I have run enough of these to know exactly what a nod is worth.”

In the old version of the workflow, the author wrote the recap that evening from memory. And that’s where things fell apart. One ownership change slipped through. Weeks later, someone built the wrong thing because they were waiting on a person who never knew they owed it.

The expert’s key insight: campaigns go sideways over the small thing nobody wrote down properly. Never the big strategic call.

I think that’s spot on. Big decisions get remembered because everyone argues about them. It’s the quiet reassignment at minute 32 that gets lost.

The five-step workflow

Here’s exactly how this industry pro ran the last kickoff. Each step is simple, and the reasoning behind each one is what makes it stick.

1. Open Granola before the call. Nothing joins the meeting as a bot. It just listens in the background and transcribes. The rationale: no awkward “recording bot has joined” moment, and nobody changes how they talk because a tool is visibly watching.
2. Brief the team like a normal human. Run the call the way you normally would. No pausing to type. No half-listening while you scribble. The rationale: your attention stays on the people, which is the whole point of a kickoff.
3. Let the summary appear on its own. When the call ends, Granola already has the summary ready, with the full transcript sitting behind it. The rationale: you’re not reconstructing anything from memory hours later. The source of truth exists the moment you hang up.
4. Send one prompt to Claude, which is connected to Granola. This is where the transcript turns into something people can act on. The rationale: a raw transcript is too long to read. A tight, structured email is what the team actually needs.
5. Read it, then send it. A quick human check, then hit send. The rationale: you stay accountable for what goes out, but the heavy lifting is already done.

Result: the follow-up went out minutes after everyone hung up. Every change, every owner, every date, written down.

The exact prompt the author used

This is the part I’d copy word for word. The creator shared the prompt they typed into Claude:

“Turn this meeting into a follow up email to the team. Every decision we made, every action with the owner named, and the date it is due. Only include what was actually said on the call. Keep it short enough to read on a phone.”

Notice how much work each line does. Let me pull it apart, because the structure is smart:

• “Every decision we made” forces the model to list outcomes, not summarize vibes.
• “Every action with the owner named” is the fix for the exact failure the author described. No ownerless tasks.
• “The date it is due” means no “soon” or “next week.” Real dates.
• “Only include what was actually said on the call” is the guardrail. It tells the model not to invent tasks or fill gaps with guesses.
• “Short enough to read on a phone” respects the fact that half your team will open this between meetings.

Why this beats the evening recap

Speed matters here more than you’d think. A recap that lands minutes after the call gets read while the meeting is still fresh. People spot mistakes and reply “actually, I think Sam owns that, not me.” A recap that lands at 9 pm gets skimmed tomorrow, if at all, and corrections never happen.

The other win is fidelity. The author isn’t relying on memory. The email comes from a transcript, and the prompt explicitly restricts it to what was said. That’s a much tighter loop than “what do I remember from this afternoon?”

How to try this yourself

A few practical tips if you want to replicate what this savvy professional did:

• Start with one recurring meeting. Weekly syncs are perfect. Low stakes, high repetition, easy to compare against your old notes.
• Tell people the call is being transcribed. Quick and courteous, and in many places it’s required.
• Keep the prompt strict. The “only what was said” line is doing the heavy lifting. Don’t drop it.
• Always read before sending. The author does. So should you. Thirty seconds of review beats a wrong owner on a task.
• Save the prompt somewhere handy. Same prompt, every meeting. Consistency is what makes the team trust the recaps.

The line that stuck with me was the closer. Having built and sold a company, the original poster says the expensive mistakes are almost never the loud ones. It’s the small unwritten stuff that costs you weeks.

If you run kickoffs, syncs, or any meeting where ownership changes hands, go read the full post on LinkedIn. It’s short, practical, and the prompt alone is worth the click.