The Meeting Ended. The Work Didn't.

AI notetakers have given me back my focus in meetings.

I don't have to quickly scribble something down while missing the next point in the discussion. I don't have to wonder afterward whether I captured the important part. I can actually listen, talk through an idea and stay in the conversation, knowing there will be a transcript and a nice clean summary waiting for me when we're done.

I love that.

But I've also started noticing what I've lost.

We have production meetings where I can get everything out of my head. We talk through projects, kick around ideas, make decisions and figure out what needs to happen next. Then the AI notetaker does its thing. It turns the conversation into a clean list of action items, often organized by the person responsible.

Meeting over. Everybody knows what they need to do.

Or at least that's what I thought.

Here's what I've witnessed.

Maybe one of the action items says, "Create a reel for Client X."

Technically, that's accurate. We did decide to create a reel.

But during the meeting, we might have spent five minutes talking about what that reel should actually be. We talked through the hook. We discussed what footage would work. Somebody threw out an idea, I built on it, and eventually we all had a pretty good picture in our heads of what we were trying to make.

The problem is that we never stopped and said, "Okay, here is the final decision."

The notetaker heard the entire conversation, but the action item became five words: Create a reel for Client X.

A few days later, it's due. Someone creates the reel. The box gets checked. And technically, the work is done.

It's just not what I had envisioned.

That's not the fault of the person doing the work. They completed the task they were given. The problem happened earlier. We mistook a record of the conversation for a handoff.

There's another piece of this I've noticed in myself.

When I used to take my own notes, I had to process what I was hearing. I had to decide what mattered enough to write down. After the meeting, those scribbled notes meant something to me because I had already translated the conversation once in my own head.

Now I can dump everything out in a meeting and move on.

That's incredibly freeing. It's also incredibly easy to think, "We talked about it, so everybody knows."

Then the AI summary arrives looking polished and complete. There are headings. There are owners. There are action items. It looks like a system working exactly as it should.

But clean isn't the same as complete.

The AI didn't misunderstand the meeting. It summarized it correctly. We just asked a summary to do the job of a handoff.

And those are two different things.

This has also made me think differently about hybrid work.

I love being able to work from anywhere. I like having a team that doesn't have to be sitting in the same building every day to get good work done. I'm not interested in giving that up.

But sometimes you need to be in the same room just working.

Not for another meeting, and not because everyone needs to sit at a desk from 8 to 5. It's because there are moments when the little conversations between the meetings matter.

You see the first cut of that reel and say, "That's not quite what we talked about." Someone asks a question while they're working on it. You pull up the footage together. The idea gets refined before anyone spends another two hours going down the wrong path.

AI can capture a conversation. It can't replace all the conversations that need to happen afterward.

I don't think the answer is simply asking AI to write better meeting notes.

The more I've thought about this, the more I think there are a few different places where the handoff can break. Sometimes we never actually made a decision. Sometimes we made one, but the action item doesn't contain enough context. Sometimes everything was clear on Monday, but by Thursday the person doing the work needs a five-minute conversation before they start.

So instead of one giant prompt, I'm experimenting with checkpoints.

This is probably the biggest one for me.

A meeting can feel productive because we talked through something for ten minutes. But talking through options isn't the same as deciding what we're doing.

Before you end the meeting, give the transcript to AI and ask:

That creates a very different list.

"Create a reel for Client X" might suddenly come back with: You discussed three possible openings for the reel, but I don't see a final decision on which one to use.

That's useful because now we can make the decision while everyone is still there instead of discovering three days later that everyone left the meeting with a slightly different version in their head.

Once the decisions are made, I want to know whether the action items can actually stand on their own.

Try this:

That last sentence matters.

I don't want AI guessing what I meant. That's exactly the problem I'm trying to solve. I want it finding the places where I haven't communicated what I meant.

Now answer those questions.

You don't need to write a novel. In fact, that's probably another trap. If every action item becomes a three-page brief, nobody is going to read it.

Once you've filled in the important gaps, ask:

Now "Create a reel for Client X" isn't just a bullet point.

It might become:

"Create a 30–45 second reel using the customer interview as the backbone. Open with her comment about why she called us, then use the installation footage underneath the rest of her story. Keep this focused on trust rather than price. First cut due Thursday. Bring me the first 10 seconds before finishing the full edit."

That's a handoff.

This might be the most important checkpoint.

Not every task needs a better written brief. Some need a conversation.

Some need someone to show you a rough version before they finish it. Some need two people sitting next to each other for twenty minutes. Some are so subjective that writing another 500 words of instructions is less efficient than opening the project together and saying, "Here's what I'm seeing."

So I also want AI to help identify those:

I really like that last part because the answer isn't more meetings.

It's figuring out where another conversation is actually cheaper than getting the work wrong.

For anything that needs more context, decide what the smallest useful checkpoint is.

Maybe it's, "Send me the hook before you edit the reel." Maybe it's, "Give me three headline options before you build the page." Maybe it's, "Let's spend ten minutes looking at the first draft together."

Or maybe it's complicated enough that the right answer really is, "Let's work on this together Thursday afternoon."

AI can help you figure out which checkpoint makes sense:

That's the system I'm trying to build.

Use AI to let me pay more attention during the meeting. Use it to catch decisions we haven't actually made and find the context that didn't make it into the handoff. Then know when to stop using AI and talk to another human.

Because I don't think the lesson here is that AI notetakers don't work. Mine works great.

The lesson is that capturing everything isn't the same as communicating everything.

And sometimes the most efficient thing you can do is spend ten minutes in the same room making sure everyone sees the same thing.

The list was never the instruction.

It was the reminder that an instruction still needs to happen.