I’m back from a few weeks off work this summer – with plenty of time for travel, reading, running and cycling.

The time away from work has been invaluable reflection time. I find the distance this gives me from the day-to-day helps me see patterns that aren’t visible when you’re in the middle of things.

One of the things I noticed this summer was that in many of the businesses I work with, it feels a lot harder to get a team aligned around a decision than it used to be.

I don’t think that’s because people are lying when they say “we’re aligned on this”, but because increasingly “aligned” describes what an AI-generated document says happened rather than what actually happened in the room or on the call.

I see this more and more in growth businesses right now and I don’t think it’s a coincidence that it’s got worse exactly as AI tools have made it effortless to produce a polished summary of anything. It’s easy to create a strategy doc, a SOP, or turn a 40-minute call into six tidy bullet points and a list of owners.

It all reads brilliantly. Everyone nods. It’s decided.

And then days or weeks later, you find out three people on the leadership team were quietly working from three different versions of what was actually decided.

There’s a writer I follow, Indy Johar, who’s been making a version of this argument about organisations generally. While I find his stuff a bit of a mental wrangle at times, this article really resonated for me when I read it while I was away.

His point (my rough summary) is that we’re getting very good at producing the appearance of shared understanding at exactly the moment our actual capacity to build shared understanding together is quietly falling away.

A well-written document can hold the appearance of agreement without any of the hard work of agreement ever happening. Nobody in the room actually tested the idea, argued with it or lived with the uncomfortable bit of it for a while. The document just arrives, sounding finished.

And because it sounds finished, everyone treats it as settled. Job done!

I’ve written before about workslop – work that looks like work but isn’t. In a way this is workslop’s more dangerous sidekick: agreement that looks like agreement but really isn’t.

It’s more dangerous because a bad piece of work usually gets found out fast when someone tries to use it for the first time. But a poorly formed decision that lacks the debate and consideration it deserved can lurk for months.

Things tend to go pop at the worst possible moment – a client crisis, a tough call on headcount or the one moment where you actually needed everyone facing the same way.

And it gets worse in the tools we use to run businesses day to day too – a couple of patterns that I’ve seen:

• Slack threads where three people drop a thumbs up isn’t the same as three people agreeing after a proper discussion where alternative points of view were debated.
• AI meeting summaries no-one reads or disputes aren’t the same as an effectuve meeting where the hard bit got said out loud and the tough work of establishing a shared viewpoint is genuinely done.

Because of this, I think you end up with superficial participation that costs the participant very little and communication that doesn’t require much attention at all.

None of this is really about whether the AI got the facts right. The facts are usually fine. What’s missing is the argument nobody had at the time – that pivotal moment in the meeting where someone says “hang on, I don’t think that’s right” and the room has to sit with it before the conversation changes direction.

So how do you tell real agreement from the AI-generated paperwork of agreement?

A few things I’ve started checking for in my work with founders and their teams:

• Ask someone to explain it without the document in front of them. If two people on your team give you meaningfully different answers about what was actually decided, you didn’t reach a proper agreement. You reached a document that looks like an agreement.
• Notice who wasn’t properly in the room. Anyone who only saw the summary – the Slack recap, the AI notes or the deck someone else built – has the artefact, not the understanding. Their nod of agreement isn’t really the same kind of “yes”. The difference matters a lot.
• Build in a moment where someone has to argue against a point that’s being agreed. Before something becomes “decided,” ask one person to make the case against it out loud. If nobody can, that’s often a sign the room agreed too easily rather than that the idea is bulletproof. This takes practice to do (it’s based on the “red teaming” concept from the military).
• Be suspicious of a write-up that’s too elegant. If a meeting had a genuinely difficult moment in it and the AI-generated summary reads like nothing happened, then the true value from the discussion has been lost. The document has been smoothed over by the AI or the person editing the AI output.

I want to be clear – none of this means slowing everything down or banning the tools that are changing how we work for the better in many ways. I use AI-assisted notes constantly and they’re genuinely useful for what they’re good at: tracing a decision, holding the memory of it or giving people something to come back and revisit later.

But they just can’t do the agreeing for you. Ever.

So a few things worth pondering this week:

• Which decision on your Notion right now would half your leadership team struggle to explain in their own words?
• How much of what you’re calling “aligned” is actually just everyone having read the same paragraph after the call?
• And what would it cost you if that gap showed up in a crisis, rather than on a quiet Tuesday when there’s still time to fix it?

This is one of the things I’ve found myself circling back to with leadership teams over and over this year.

I’m happy to talk through where your own version of it might be hiding, if that’d help.