My Action Items Kept Falling Through the Cracks

The problem was never that I forgot what happened in meetings.

It was that I left meetings with a vague sense of what I had agreed to do, carried that vague sense through the rest of the teaching day, and then rediscovered it three days later when someone followed up and I had to reconstruct from memory whether I had said I would do the thing or only that I thought it was a good idea.

Those two things feel different in the meeting. They look identical in your memory forty eight hours later.

AI meeting summarizers get talked about mostly as note-taking tools. That framing undersells the specific problem they solve for teachers, which is not note-taking. Most teachers are adequate note-takers. The problem is action item clarity. What exactly was decided. Who exactly agreed to do what. By when. In a staff meeting where twelve people are talking about four different agenda items and some decisions are made explicitly and some are made by implication, those details dissolve faster than you expect.

This piece covers three free tools that address that specific problem in different ways, and one situation where none of them should be used regardless of convenience.

Why Teacher Meetings Are Harder to Summarise Than Most

Before the tools, the honest context. Teacher meetings are not like corporate meetings where an agenda is followed, decisions are clearly stated, and someone types minutes in real time.

A typical department meeting moves between planned agenda items and unplanned conversations, includes decisions made by implication rather than explicit agreement, references context from previous meetings that not everyone remembers, and occasionally produces commitments that the person making them does not fully register as commitments because they were phrased as suggestions.

An AI meeting summarizer does not fix any of those dynamics. It captures what was said, not what was meant. If your department head says "it would be good if someone could pull the data before next week" in a way that everyone in the room understands means you should pull the data, the summarizer captures the sentence. Whether it interprets it as an action item assigned to you depends on how directly the language matched the tool's action item detection patterns.

This matters because the gap between what a summarizer captures and what actually needs to happen is the gap where action items still fall through. The tools help. They do not eliminate the problem entirely.

Tool One: Otter — Best Free Option for Automatic Action Item Extraction

Otter is the most established free AI meeting transcription tool available to teachers in July 2026. The free tier includes three hundred minutes of transcription per month, automatic summary generation after each meeting, and action item extraction pulled from the transcript.

The action item extraction is what matters most for the problem this article is about. Otter scans the transcript for language patterns associated with commitments and tasks and pulls them into a separate list. Statements like "I will send that by Thursday," "can you follow up with the district office," and "let's make sure someone checks the schedule" surface as action items with the speaker identified where it can determine who said what.

The accuracy is good on direct commitment language and worse on implied commitments. "It would be good if someone pulled the data" does not consistently surface as an action item because the language is indirect. "I will pull the data before next week" almost always does.

The practical implication: Otter works best in meetings where decisions and commitments are stated directly. It is less reliable in meetings where the culture involves indirect communication about who is responsible for what. If your school's meeting culture tends toward direct language, Otter's action item extraction is genuinely useful. If it tends toward implication and suggestion, you will still need to do some manual review to catch what Otter missed.

Free tier: three hundred minutes per month. For a teacher attending roughly four one-hour meetings per week, this covers about one week of meetings per month before hitting the ceiling. Selective use — prioritising the meetings where follow-through matters most — is more sustainable than trying to run Otter on everything.

One limitation worth stating plainly: Otter's free tier processes audio through external servers. Do not use it in meetings that involve identifiable student information, IEP discussions, parent concerns, disciplinary matters, or personnel topics. The data handling implications of recording those conversations with a free commercial tool are not appropriate regardless of how useful the transcription would be.

Tool Two: Claude via Pasted Notes — Best for Meetings Where You Cannot Record

Some meetings cannot be recorded. The participant prefers not to be recorded. The school culture discourages it. The content is sensitive enough that recording feels inappropriate even for your own use. In those situations, the recording-based tools are not an option but the action item problem is still real.

The workflow that addresses this: take rough notes during the meeting as you normally would. Immediately after the meeting ends, open Claude and paste your notes with this prompt:

"These are my rough notes from a staff meeting. Please pull out a clear list of action items — things that need to be done, who was mentioned in connection with each one, and any deadline that was stated. Also note any decisions that were made that I should remember but that are not action items. Keep the language close to what my notes say rather than interpreting or inferring."

The instruction to keep language close to the notes rather than inferring is important. Without it, Claude occasionally promotes an implied suggestion into a firm commitment or assigns an action item to someone based on context that was not clearly stated in the notes. The instruction tells it to surface what is there rather than interpret what was meant.

This workflow does not require recording. It does not require a separate app. It requires rough notes, which you were already taking, and two minutes after the meeting to paste them and read the output.

The limitation is obvious: the quality of what Claude extracts depends on the quality of your rough notes. If your notes are incomplete because the meeting moved fast, the output is incomplete. But the act of reading the extracted action items immediately after the meeting, while the conversation is still fresh, lets you catch gaps and add what you remember before it fades.

Tool Three: Google Meet With Gemini Summaries — Best for Meetings That Happen in Google Meet

For teachers whose staff and team meetings happen in Google Meet, Gemini's built in summary feature removes the step of running a separate tool. After a Google Meet session with Gemini summarization enabled, a summary and action item list are sent automatically to participants.

The quality is adequate for straightforward meetings. The action item extraction is less precise than Otter's — Gemini is more likely to list discussion topics as action items and more likely to miss indirect commitments. For meetings with clear, explicit decisions and directly stated follow-up tasks, the automatic summary is useful. For meetings where the important decisions were implied, the Gemini summary requires more manual review.

The genuine advantage is zero friction. Nothing to set up before the meeting. Nothing to open afterward. The summary arrives in your inbox. For teachers who would not consistently use a separate tool because of setup friction, a worse summary that arrives automatically beats a better summary that requires opening an app.

Availability: depends on your school's Google Workspace for Education configuration and whether your account has Gemini features enabled. Verify with your school's IT department before planning to use it.

The Meeting Type Where None of These Should Be Used

Any meeting involving identifiable student information.

IEP meetings, student support team discussions, behavioral review meetings, parent-teacher conferences, disciplinary conversations, any discussion where a specific student's academic performance, disability status, behavioral history, or family situation is discussed — these meetings should not be recorded with a free commercial AI tool and the transcript should not be pasted into an AI platform for processing.

Student information is protected by FERPA in the United States, UK GDPR in the United Kingdom, and equivalent protections in most other countries. Free commercial tools are not typically covered by the data processing agreements that make using student information with third party AI tools compliant.

Take written notes in those meetings. Keep them in your school's secure system. The action item problem in an IEP meeting is better solved by having a clear meeting facilitator who confirms action items aloud before the meeting closes than by running AI transcription.

The Habit That Makes Any of These Tools More Useful

The summarizer is only as useful as what you do with the summary.

Teachers who consistently follow through on meeting commitments after using these tools have one habit in common: they read the action item list within ten minutes of the meeting ending and move each item to wherever they track their to-do list. Not later that day. Within ten minutes, while the meeting context is fresh enough that the action item makes sense.

An action item list that sits in an Otter summary or a Claude response and gets read three days later is only marginally better than a meeting note that says "important!!" with two exclamation marks and no other context. The summarizer creates the list. What you do with the list in the first ten minutes is what determines whether the action item actually happens.

That ten minute transfer is not something a tool can do for you. It is the habit that makes the tool worth using.