Meeting Action Item Extractor

Extract action items from meeting notes without inventing them

Most meeting tools give you a summary. A summary tells you what the meeting sounded like. It rarely tells you what changed. Two days later you still can’t answer the only questions that matter: what did we decide, who owes what, and what did we leave hanging?

This free meeting action item extractor takes pasted notes, call notes, or a rough AI-generated transcript and sorts them into five buckets — decisions, action items, open questions, follow-ups, and uncertain items — so the output is a work list rather than a paragraph.

Why a summary and an extraction are not the same thing

A summarizer optimises for readability. It compresses, smooths, and fills gaps so the text flows. That is exactly the wrong behaviour for meeting follow-through, because the gaps are the useful part. When a summary says “Marcus will update the onboarding emails,” you cannot tell whether Marcus committed to it, whether a date was agreed, or whether the model inferred it because Marcus mentioned emails.

An extraction optimises for accountability. It keeps unclear things unclear on purpose, and shows you the line it drew each item from.

Example: how ambiguous ownership and deadlines are handled

Take these three lines from a real-feeling set of notes:

“I’ll ship the toggle change by Thursday.”
“I’ll take a look at the onboarding emails at some point.”
“Someone needs to update the copy under the plan cards too.”

A generic summarizer flattens all three into “the team will update pricing content.” Here they come out separated:

That third row is the whole point. It is probably a task. It is not yet a commitment, and promoting it to one would put a false entry in your tracker.

What each section means

• Decisions — settled outcomes. A preference nobody agreed to is not a decision.
• Action items — concrete tasks, with owner, due date and any status context that was actually stated.
• Open questions — raised, discussed, not resolved.
• Follow-ups — needs another conversation, a confirmation, or someone outside the room.
• Uncertain / inferred — may be an action, but the notes don’t support stating it as fact.

How to get a better extraction

• Keep speaker names or initials in the transcript — ownership depends on them.
• Paste the raw notes rather than a cleaned-up version; cleanup usually deletes the evidence.
• Add the meeting date if people said “next Friday” or “end of month”.
• Run long workshops in sections rather than one huge paste.
• Read the uncertain section first — it is where the follow-up email usually needs to start.

Privacy: what happens to the text you paste

The notes you paste are sent to an AI model to produce the extraction and returned to your browser. They are not written to our database and are not linked to an account. The only thing stored is a hashed value derived from your IP address, used to rate limit the free tool. We can’t make promises about the model provider’s own handling of transient request data, so treat this the way you would treat any third-party AI service: don’t paste material you aren’t allowed to send outside your organisation.

Frequently asked questions

How is this different from a meeting summary?

A summary compresses what was said. An extractor pulls out the parts that change what happens next: what was decided, who owes what, what is still unanswered. A summary can read well and still leave you with no task list. This tool sorts the notes into decisions, action items, open questions, follow-ups and uncertain items instead of producing paragraphs.

What happens if nobody was assigned a task?

The owner field says "Owner not stated" rather than guessing. Same for deadlines: if no date was said, the tool returns "No due date stated". Guessing an owner is how a task quietly becomes nobody's job.

Does it show where each item came from?

Where practical, each decision and action item includes a short verbatim excerpt from your notes. That lets you check the extraction in a few seconds rather than re-reading the whole transcript.

What are uncertain or inferred items?

Things that look like they might be an action or a decision but were not stated clearly enough to present as fact — a suggestion nobody confirmed, or a task with no committed doer. They are kept in a separate section so you can decide, not the model.

What happens to the text I paste?

Your notes are sent to an AI model to produce the extraction and returned to your browser. They are not saved to a database or attached to an account. We store only a hashed value derived from your IP address so the free tool can be rate limited. Do not paste anything you are not permitted to send to a third-party AI provider.

Is there a length limit?

Yes — up to 20,000 characters per run, which covers most single meetings. For a long workshop, run it in sections.

Does it connect to Zoom, Teams or my project tracker?

No. This version is deliberately paste-and-process. You copy your notes in and copy structured results out as text or Markdown.

Keep going

Once you have the action list, the follow-up email is the next job. Use the AI Assignment Builder to turn a vague item into a properly scoped brief, the Client Onboarding Email Generator when the meeting kicked off new client work, or browse the full free AI tools and prompt library.