A couple of weeks ago I ran a strategy offsite for a PE-backed software business. 16 executives in the room, all of whom I had interviewed individually beforehand — by the time we started, I knew what needed to be said that day, and roughly who was sitting on it.

But for the first two hours, everyone played along. Things were going to plan.

I watched two people start to say something and stop, the sentence dying halfway, rolling their eyes instead. Someone opened her notes from our one-on-one, read them, and closed the folder again. Nobody was following the plan in their head. They were waiting to see who would go first.

It broke in the second coffee break, in the kitchen. The CFO was rinsing her mug at the sink, half turned away from the group, when she said it, almost to the tap. "We all know this isn't working." Spoons stopped mid-stir.

I thought: “Now we’re talking, finally.”

The challenges broke out into the open: No product-market fit, eroded over years of catering to a legacy cohort of clients. Knee-deep in technical debt, so far behind that the market is running away from them, and no version of the next two years that doesn't involve rebuilding the product and engineering org completely.

I had heard all of it in the one-on-ones. Every person in that room had known some part of it for months. The agenda we had all agreed on was the only place it couldn't be said.

An agenda keeps people in character. That's what the first one is for. I build it so the team can arrive and feel safe, and then I spend the morning pushing against it.

Because secretly, I carry two designs into every offsite.

The first one is the one everybody has seen. The second has no slides and only becomes usable after the break. The work is the handover between them. The issues that matter can't go on an agenda, because writing them down is already saying them out loud.

It's a bit like surfing: it's only fun once the wave starts to break.

And it rarely breaks in the session. It breaks at a sink, or over lunch, in the twenty minutes when nobody is obliged to represent their function.

Which raises the question of why you'd put 16 people in a room for two days at all. Everything they contributed in those first two hours could have been collected in a shared document, on their own time, at a fraction of the cost. Position papers, functional updates, proposals — that is async work, and doing it in a room doesn't make it better, just slower. I run it in the room anyway, because a document doesn't break. Asynch collaboration has no kitchen for the CFO to rant.

Afterwards I run an analysis on the transcript, and what I'm looking for is the nature of every contribution. There are three:

Advancing is the part that could have been a document. Inquiring and aligning are the two that require the room — and before the breaking point there is almost none of either. Everyone advances from inside their own function, contributing exactly what their role obliges them to contribute. Afterwards, if the team can hold what was said without finger-pointing and with some curiosity about how to solve it, the distribution moves into something far more balanced. If the team collapses into conflict instead, it stays just as lopsided, except now everyone is defending rather than reporting.

Everyone in that room knew the break had happened. You don't need a transcript for that. What you need it for is the day after: whether the team actually came unstuck, or whether they said the true thing once and then went back to their functions. That difference doesn't announce itself.

Here's the cheapest AI-based team dynamics diagnostic you can get:

Take your last leadership meeting, run the transcript through whichever model you use, and have it sort every contribution into advancing, inquiring and aligning. Have the analysis break out individuals contribution style as well. Then, look at the ratio across the room, and at whether anybody's questions went anywhere near a part of the business that isn't theirs.

Then share the results with the team. Send them the distribution as it came out, without your interpretation, and let them go first.

The conversation that follows is the one you couldn't have scheduled.

Yours,