Two ears, one mouth wasn't the problem

A few months ago I started recording my meetings and reviewing them with Claude. For some meetings I did my prep with Claude and got its advice combined with my goals and then I did a review after.

I don’t know why, but I was caught by surprise.

I like to think I am good at holding a conversation and can communicate well. I also knew I tended to spend time talking about things for too long - even though I know that you should listen more than you speak.

Two ears, one mouth - that’s the ratio of listening to talking to aim for. I tend to be about 60:40 in favour of listening. So all good…

Except no.

Again and again, there were moments where a slightly different framing of what I had just said might have shifted the conversation in a better direction and I only ever spotted it afterwards, never in the meeting.

I was leaning hard on personal anecdotes, the stories that build warmth and make a call feel human, and giving them more time than the actual goal of the call could afford.

Every minute spent building warmth is a minute not spent moving towards the outcome you are there for, and I had that balance wrong more often than I would have guessed.

This isn’t just for sales calls.

Surgeons have been doing a version of this since 2010 - Dr Teodor Grantcharov built a “black box” for operating rooms, borrowed straight from aviation, and the residents who reviewed their own footage improved more than the ones who did not. He has said that the first time he watched a tape of himself operating, he wanted to throw it out the window.

Teachers who submit classroom video for coaching see measurable gains in what their students learn, not because a coach tells them what to change, but because they finally see what students were reacting to in the room. Therapists in training get more useful supervision from a recording of a session than from their own account of it, for the same reason.

Yes, knowing you are being recorded changes how you behave - one survey found 84% of people alter their speech once they know AI is listening. That is real. But it does not last.

Reactivity, the term researchers use for behaving differently because you know you are being observed, is one of the most consistently replicated findings in the field: it fades with exposure, because it is hard to sustain unnatural behaviour for very long.

Studies using wearable audio recorders that sample people’s days at random - even ones worn openly, with a visible microphone - found participants adapted within days, reporting low discomfort and acting close to normally once the novelty wore off. The adjustment is real on day one. By the second week, most people forget the recorder is there.

I suspect we are getting the same situation now with online meetings - we know they are being recorded and forget about it.

Regardless, none of this is really about AI. The cost and ease of reviewing has collapsed. For those who want to improve continuously, it is a habit you will not drop.

This AI-powered “black box” could make surgery safer - MIT Technology Review - the detail worth your time: the box does not just film the surgeon, it captures room temperature, noise levels, and device data at the same time, because errors turn out to be as much about the environment as the individual. Worth reading if you have ever assumed a mistake was purely a skills problem.

Pick one type of conversation you have every week - a client call, a 1:1, a pitch - and record it, with the other person’s knowledge. Before you listen back, write down what you think you achieved and how you got there.

Then check the recording for two things specifically: where you could have framed something better in the moment, and how much time went to building rapport versus moving towards your actual goal. Or just ask Claude and see if you agree.