Summarize meetings with AI

The reason meeting notes never get written is that writing them is tedious and everyone assumes someone else will. AI removes the excuse: paste a transcript, get a structured summary with action items, in about ten minutes. But there's a right and wrong way to prompt it. Ask for "a summary" and you get a bland paragraph nobody acts on. Ask for the *specific structure* a team actually uses — decisions, action items, open questions — and you get notes that drive follow-through. This workflow does the second, for free.

The structure is the entire value. A meeting summary that doesn't separate "what we decided" from "who's doing what" is just a shorter transcript.

What you'll walk away with

• A structured summary: Key Decisions, Action Items (with owners), and Open Questions
• Action items converted into trackable to-dos, not buried in prose
• A filed, dated note you can actually find again later
• A ten-minute habit that replaces "wait, what did we agree on?"

Step by step, with the judgment that matters

1. Summarize with ChatGPT — dictate the structure. Paste the transcript and ask specifically for three sections: Key Decisions, Action Items (with owner if mentioned), and Open Questions. That last section is the one people forget and the most useful — it captures the "we'll figure this out later" moments that otherwise vanish. Owners matter: an action item without a name attached is an action item nobody does.

2. File it with Notion AI. Paste the summary into a Notion page, convert each action item into a checkbox to-do, and tag it with the meeting date. This is what turns a summary from a document you never reopen into a live task list. If a video-call tool already dumps transcripts into Notion, this step chains right onto it.

A real run

After a 45-minute planning call, you paste the auto-generated transcript into ChatGPT and get three clean sections — four decisions, six action items (three with named owners, three you assign yourself), and two open questions about budget. In Notion you turn the six items into dated checkboxes. Ten minutes after the call ends, everyone knows what they own, and the two open questions are on record instead of forgotten.

Where it goes wrong

• "Where do I get the transcript?" Zoom, Google Meet, and Teams all offer auto-transcription; or run a recording through Descript or Otter.ai first.
• The summary missed a key point. Long transcripts can exceed a tool's limits. Split a very long meeting into two chunks and summarize each.
• Action items have no owners. The transcript didn't name them, so the model can't invent them. Assign owners yourself during the Notion step — an unowned task is a dropped task.

The honest cost

Free at the core. ChatGPT's free tier handles the summarizing. Notion AI is a paid add-on (from ~$8/member/month) — if you'd rather stay free, skip it and paste the summary into plain free Notion, turning the action items into checkboxes manually.