Action Item Extractor

Extract actionable tasks and next steps from LLM-generated notes or reports.

Ad placeholder (leaderboard)

Turn LLM notes into a clean action list

LLMs are great at summarising a meeting or drafting a report — but the next steps end up buried in prose. This tool reads that prose and pulls out the action items, separating the task, the owner, and the deadline into a tidy checklist you can paste straight into a tracker. It runs entirely in your browser.

How it works

Each sentence is scored for “action-ness”:

  • Modal / intent verbswill, must, should, need to, plan to, going to.
  • Assignment languageassign, owner, responsible, action:, todo, follow up.
  • Imperative openings — a sentence that starts with a bare verb (Send …, Review …, Schedule …).
  • Date signalsby Friday, due 12 March, before EOD — these boost the score because a dated sentence is almost always a commitment.

Sentences above the sensitivity threshold become tasks. The extractor then runs a second pass to lift out a likely owner (@name, assigned to X, a leading proper noun) and a deadline (relative or absolute date phrases), shown only when confidently found.

Tips and notes

  • Set the model up for success: prompts like “List next steps as ‘Owner — task — due date’” make extraction near-perfect.
  • If you get false positives, raise the sensitivity; if real tasks are missed, lower it.
  • The Markdown output uses - [ ] checkboxes, so it renders as an interactive task list in GitHub, Notion, Obsidian and most trackers.
Ad placeholder (rectangle)