The expensive part of a meeting is the follow-up nobody owns. Notes say "Lina can look into reminder emails" with no date, or "someone should check the invoice template" with no name, and a week later nobody remembers.
This workflow reads meeting notes and produces a task table plus a short follow-up email draft. It was tested on four meetings on October 26, 2026: a sales sync, an ops check-in, an empty note and a meeting that was only small talk.
Who it is for
A team that holds meetings and loses the follow-ups: who does what by when.
How it works, step by step
- Rejects meetings with empty notes before the model.
- A small local model lists tasks, owners and due WORDS exactly as written.
- Code checks owners against the team and attendee list, converts weekday words to real dates using the meeting date, and flags vague or missing dates and owners.
- Writes action_items.csv with status ready or needs-human.
- A second local call drafts a short follow-up email from the ready items only; code blocks drafts that add a date not in the list.
The model copies tasks, owner words and due words. Code then checks every owner against the team and attendee lists, converts weekday words into real dates using the meeting date, and flags anything vague. The follow-up draft is only built from the ready items, and a check blocks any date that is not in the task table.
How it runs: the execution strategy
It has two triggers: a schedule (weekdays at 18:00, after the day’s invoices have arrived) and a manual trigger for testing. Each run checks the saved invoice index first, so anything already recorded is reported as a duplicate, then validates and routes every invoice to the ledger, the approval list or the review list. It has not been connected to a live mailbox; the runs described here were run by hand.
The workflow on the n8n canvas
Screenshots taken in n8n right after a run. A green tick means the step finished, and the number on each line is how many items passed through it. Click an image to see it larger.


What went wrong while building it, in order
This list is copied from the build log kept while the workflow was made.
- First run, 4 meetings: 8 rows. Ready 2, needs a human 4, no actions found 1, rejected 1 (empty notes never reached the model). 80 s total, extraction 52 s, follow-up draft 8 s.
- What the checks caught (all by code, not by the model): "Wednesday" turned into 2026-10-28 using the meeting date; "starting tomorrow", "no date" and "next week" were flagged instead of guessed; "Priya" was not on the team or attendee list so the task became UNASSIGNED (Priya?); an unnamed owner became UNASSIGNED.
- Quality notes, not tuned: one task came out as "takes the backup test" (the model copied the sentence fragment) and the follow-up draft only lists the ready items. A person edits both before anything goes out.
- A sentence that was not an action (the demo slot on Friday at 11) was correctly left out; the small-talk meeting produced "no action items found" instead of invented tasks.
- T2 wrong model name: every meeting becomes a needs-human row and the follow-up draft is blocked with an explanation instead of an empty or invented email.
What we learned
- Refusing to guess is a feature. "Starting tomorrow", "no date" and "next week" were flagged as needs-human instead of being turned into plausible-looking dates. A name that was not on the team list (Priya) became "UNASSIGNED (Priya?)".
- Not everything said in a meeting is an action. The demo slot "Friday at 11" was correctly left out, and the small-talk meeting returned "no action items found".
- Quality limits showed up too. One task came out as a sentence fragment ("takes the backup test"). We recorded it rather than hiding it: a person edits the list before anything goes out.
What the finished run produced

This is the real output file from a clean first run, with every saved file removed beforehand. The whole run took 78 seconds on a CPU-only laptop, including about 15 seconds of n8n start-up.
# Meeting run summary Rows: 8 | ready 2 | needs-human 4 | no-actions 1 | rejected 1 - [ready] mtg1: Send revised quote to Riverside client | owner Sana | due 2026-10-28 <- weekday "Wednesday" resolved from meeting date - [needs-human] mtg1: Update lead tracker after every call | owner Omar | due - <- vague due date "starting tomorrow" (ask for a real date) - [needs-human] mtg1: Look into reminder emails | owner Lina | due - <- no due date - [needs-human] mtg1: Check invoice template | owner UNASSIGNED | due - <- no owner named; no due date - [ready] mtg2: takes the backup test | owner Dev | due 2026-11-02 - [needs-human] mtg2: review the contract wording | owner UNASSIGNED (Priya?) | due - <- owner "Priya" is not on the team or attendee list; vague due date "next week" (ask for a real date) - [no-actions] mtg4: (no action items found) | owner - | due - <- model found no actions; check the notes if this seems wrong - [rejected] mtg3: (skipped) | owner - | due - <- no notes text
Node-by-node trace of that run
| Node | Status | ms | Items out |
|---|---|---|---|
| Manual Trigger | success | 5 | [1] |
| Settings (edit me) | success | 7 | [1] |
| Preflight: local model server | success | 181 | [1] |
| Server up? | success | 15 | [1, 0] |
| Input: meeting notes (replace with transcript or notes file) | success | 1716 | [4] |
| Prepare each meeting | success | 8 | [4] |
| Has notes? | success | 8 | [3, 1] |
| Build prompt | success | 6 | [3] |
| Local model: extract actions | success | 44829 | [3] |
| Normalise owners and dates | success | 24 | [8] |
| Build task list and follow-up prompt | success | 9 | [1] |
| Local model: follow-up email draft | success | 10444 | [1] |
| Check draft and assemble outputs | success | 12 | [3] |
| Text to file | success | 26 | [3] |
| Write outputs | success | 29 | [3] |
Limits, stated plainly
Four meetings is a small test. The follow-up draft lists only ready items, so a person still has to deal with the needs-human rows. Nothing is emailed by the workflow. Everything above was tested on a laptop . We make no claim about results on your data.
What it costs
Software: nothing. n8n Community Edition is free to self-host and the AI model runs locally with a free model runner such as Ollama. You need a laptop (8 GB of memory minimum, 16 GB is comfortable, no graphics card), an internet connection for installing, and about 20 minutes the first time. No cloud AI account, no API key and no subscription.
Want the workflow file?
The workflow file and a plain-English guide exist for every workflow in this series. If you would like this one, or you have a repeated task in your own business you would like checked for fit, tell us about it. We will say honestly whether a workflow like this could take it off your plate, and we will not promise results we cannot show.
How this article was produced
This article was drafted with AI assistance from the build log, test traces and output files of a workflow built and run on the site owner's own laptop, then published by the site owner, who is responsible for it. The workflow was built and tested by an AI assistant under the owner's direction. If you find an error, tell us. Our standards are described in the editorial policy.
What is and is not verified
Verified: the build-log events, the node trace and the output shown come from real runs on October 26 and 27, 2026. Not verified: how the workflow behaves on real business data, with other models, or at larger volume. The node diagram is generated from the workflow file and is not an n8n screenshot.
Sources
Related: Search Console Says 43 Pages Are Not Indexed: What We Tested and What We Found and Moving to a custom Blogger theme: 9 errors we hit.
About the author
I'm Junaid. I'm a sales professional with more than seven years across several industries, and I have no coding background. I use AI to build and test tools like this and I write down exactly what happened, mistakes included. If a process in your business is eating your time, tell me about it and we can look at whether AI could take some of the load.