I Sat in a Meeting With More AI Notetakers Than Humans

Last updated: 2026-08-30

I was testing multiple AI meeting assistants at once. Other people on the call had brought theirs too. I looked at the participant list and there were more AI notetakers than humans.

Funny snapshot of where work is headed. Everyone wants the meeting captured. The visible bots still distract people, and they create resistance.

If your inbox is full of transcripts and summaries, you already know what happens next. They pile up. They’re searchable. You almost never go back and do anything with them. Brooks calls this the meeting notes graveyard. He uses recordings mostly to find information later. My inbox fills with summaries I skim or ignore.

Searchability is useful. It’s also a reactive, level-one use of these tools. You captured the meeting. Then you treated the capture as the job.

The discussion in most meetings is fine. What fails is converting that discussion into accountable action. Brooks has watched the corporate version of this for years: the same issue gets re-discussed two months later because nothing left the meeting in an actionable format.

So if you’re going to use an AI meeting assistant, use it on purpose.

Ask before you hit record.

Brooks’s etiquette rule is simple. Ask first. Use this:

“Are you okay if I use an AI notetaker for this call?”

Copy it. Paste it. Especially on sensitive one-on-ones, medical, legal, coaching, or confidential conversations. Brooks doesn’t mind being recorded. He still wants to be asked. People also behave differently when an assistant is present. Skip the ask and you get a worse meeting, plus a room that’s already annoyed with you.

Choose how you capture. Don’t stack bots by default.

There’s a real tradeoff between invisible local recording tools and visible meeting bots.

A bot that joins the call is obvious. Everyone can see it. That can be honest, and it can also be the reason people tense up. I’ve seen that happen when too many assistants show up at once.

A tool that records computer audio never joins the call, so the meeting has less friction. Local transcription can be better for privacy. Speaker identification may be weaker. You still have legal and ethical consent to think about. Invisible is not the same as “you don’t need permission.”

Then there’s the recorder already built into the meeting platform. No extra guest. Fewer surprises.

Before your next call, pick one: a bot-based assistant, a local recorder, or the built-in option. Adding another bot because everyone else did is how you end up with a participant list that looks like a robot convention.

Put the output where work already happens.

A strong assistant updates the systems people actually use. Leave action items inside a transcript and they die there. Connect the assistant to a task manager, a CRM, a project management app, or team chat so the output lands where people already execute. Some tools can trigger that through built-in integrations or an automation platform. A summary-only tool cannot.

Run every meeting through this checklist:

1. Transcript
2. Summary
3. Actions
4. System update
5. Human review

That last step is the one people skip. The assistant drafts. You confirm. If a task lands on the wrong person, that’s on the review, not on the recording.

Start with the admin you keep dodging.

CRM updates are a high-ROI use case because sales and client-facing people avoid the admin even when it helps them. CRM hygiene has been a long-running pain point in software and sales teams. Brooks has seen that one hang around. After a call, an assistant can detect new contact details, meeting notes, buying timelines, and follow-up commitments, then update the CRM or create a reminder.

I’ve used this in straightforward ways: update a contact’s location, add call notes, create a 30-day follow-up task. Stop relying on memory after client calls. Memory is how the follow-up never happens.

There’s a second use if you create anything from conversations. Transcripts can become a content and customer-insight engine. Capture testimonials, customer pain points, research insights, and personal stories into a searchable content bank. Later that bank can support newsletters, podcast outlines, and posts. Someone mentions an interesting story or a testimonial in passing. You don’t need a separate notebook. Capture it, store it, find it when you actually sit down to write. You’re already having the conversations. Build the bank from them.

Make the meeting produce owners and dates.

After you hang up, you should be able to answer one question: who promised what, by when? Push those tasks into a project management system. Decisions, owners, deadlines, automatic task capture. If an item has no owner and no date, it will come back in two months as a “quick recap” of a problem you already talked through.

Do this on your next call. Before you join, send the consent line. After you hang up, don’t file the summary and call it done. Push one real action into the system where you already work, then review it. One meeting. One action that leaves the graveyard.