AI Now Handles the Meeting Notes We Used to Write

Every team I’ve worked with runs the same expensive experiment: sit smart people in a room, have one of them half-listen while typing notes, then wonder three weeks later why nobody agrees on what was actually decided. We ran that experiment for years. Then we stopped.

At what., we’ve replaced manual meeting notes with AI almost entirely. Not because it’s trendy, but because the old way was quietly costing us hours every week and giving us worse documentation to show for it. Here’s what we changed, what surprised us, and how to get started without overthinking it.

The tax you pay twice for every meeting

Manual note-taking has a math problem. You pay for the meeting once by attending it. Then you pay again reconstructing decisions, chasing down action items, and re-explaining what was agreed to people who weren’t fully listening because they were busy typing.

Industry research backs this up: without notes, roughly 70% of meeting decisions are forgotten within 24 hours. Notes taken by different people, in different formats, with different priorities, make that worse, not better. You’re either fully in the conversation or accurately capturing it. Rarely both.

In Switzerland, where labour costs are high, this isn’t a minor inefficiency, it’s real money. A couple of hours saved per person per week compounds fast across a team. If you’re already looking at where AI automation can take repetitive work off your team’s plate, meeting notes are one of the easiest places to start.

What happens when AI sits in on your meetings

AI note-takers join the call, transcribe everything in real time, tell speakers apart, and pull decisions and action items into a structured summary, not one person’s rushed interpretation of what happened.

We started with Gemini because it’s built directly into Google Meet and needed almost no setup. It worked well until we hit a wall: German, and especially Swiss German, isn’t recognised reliably by most mainstream tools. That pushed us to Optiverse, a Swiss transcription tool that handles it properly and is also more compliant with local rules than most US-based alternatives.

Today we have a transcriber invited to almost every meeting by default, with an easy opt-out for anything personal or sensitive. Optiverse emails a summary right after the call ends: decisions, open topics, next steps. Nobody has to “own” the notes anymore.

This only works if the tool actually fits into how you already work. Many “AI problems” clients bring to us turn out to be tools integration problems in disguise: the AI part works fine, it’s the handoff to the rest of the stack that’s broken.

What changed once we stopped taking notes ourselves

Before, someone always had to own the notes. That person was half-present, follow-up was inconsistent, and decisions got disputed weeks later because nobody had a shared reference.

After, everyone stays in the conversation. Summaries show up automatically. Action items have owners attached, and searchable history means “what did we agree on in March?” takes ten seconds instead of a Slack archaeology dig.

We can now confidently skip people from meetings who don’t need to be there live, because they’ll get a reliable recap either way. And the closing five-minute recap, where the host confirms decisions out loud, actually gets captured properly instead of evaporating the moment the call ends.

One honest caveat: AI still mis-tags action items occasionally, or misses domain-specific terminology. Treat it as a first draft, not gospel. A quick human review of anything critical is still worth the two minutes it takes.

How to get started without overthinking it

You don’t need a company-wide rollout to start. Pick one recurring meeting and test it there.

• If your team speaks English, just use whatever note-taker ships with your meeting platform. Google’s or Microsoft’s built-in options are good enough for most teams.
• If you’re working in German or Swiss German, a specialised tool like Optiverse will save you a lot of frustration since mainstream transcription still struggles with dialect.
• For in-person meetings, we’ve found that having everyone say their name at the start helps Optiverse tell speakers apart. Results can vary by tool and setup, so it’s worth testing with whatever transcription tool you use before relying on it.
• Agree as a team on what the AI notes are actually for: a source of truth for decisions and action items, not a transcript everyone has to read word for word.

If you’re a Swiss team, pay attention to where your data is processed and stored. It matters more for meeting content than most people assume, and it’s worth checking before you standardise on a tool. We cover this in more depth in our piece on data privacy and security for AI systems.

Also relevant: if you want more quick wins like this one, read The 15-Minute AI Workflow Setup That Saves 2 Hours a Week for two more workflows your team can set up just as fast.

From meeting notes to automated follow-up

We’ve since taken this further by building and connecting several AI tools and systems together, so that once a meeting ends, we can simply say “follow up” and the right thing happens. It finds the right meeting, checks a few predefined rules, and decides whether we need a Slack recap for the team, a summary email for a client, or a new ticket in our project tracker. What used to take 30 to 60 minutes now takes 10 to 15, and we still review and approve everything manually before it goes out.

Manual meeting notes are a tax your team pays twice. AI removes that tax without lowering the quality of what gets captured. In most cases, it improves it. The goal was never to automate the meeting itself. It’s to make the time spent in it actually count.

If manual admin work like this is quietly draining your team’s time, it’s usually not an isolated problem. Explore how AI automation can take repetitive work off your plate. Meeting notes are a great first place to start.