Half your team uses AI in secret. The other half doesn't even try.

One of these halves you will find out about the day something goes wrong. The other you will never find out about, because it is made of what did not happen.

In a team where nobody has agreed on anything about AI, people split into two groups — and neither is the one the owner pictures.

Those who use it in secret do it their own way, and those who do not use it stop producing while they wait for someone to say they may.

Notice that neither group is acting in bad faith. The first wants to deliver better and faster, and worked it out alone because there was no rule. The second is being cautious in exactly the way they were taught to be.

The half you will find out about

This is the one that usually frightens people, and the easiest to picture: someone pastes an entire contract into a tool to get a head start on the analysis, or drops the customer spreadsheet into a free personal account.

What makes that dangerous is not the tool. It is that the decision about what may leave the company is being made, one instance at a time, by whoever is in a hurry — with no written criterion, no record, and nobody aware that a decision was made at all.

And there is an uncomfortable detail: the person doing it is usually the most dedicated one. The careless employee uses no AI at all. They do it the old way, and show up in no risk report.

The half you will never find out about

This one is bigger, and invisible by construction.

It is the person who would open the tool and solve in twenty minutes what will take two days — and does not open it, because nobody said they could. That is not laziness. It is the correct reading of a company that has not spoken: in silence, the safe move is to do nothing.

That loss never becomes an incident. It does not make the news, it does not enter a report, and nobody measures it. It only shows up sideways, when a competitor delivers in two days what your team takes two weeks to do — and even then nobody connects the two.

That is why almost every company thinks its AI problem is the first group. The first group is the one that shows.

What the two halves have in common

They look like opposites and share one cause: nobody said where the line is.

It is not a failure of discipline — it is an absence. As long as it lasts, each person will fill the gap with their own judgement, and the two groups will go on existing in parallel, never speaking. Including this: whoever found a good way to use AI shares it with no one, because they are not sure they should be using it at all.

Look at the size of the waste. The company is paying for both losses at once — the risk from those who use it without criteria and the productivity of those who do not use it — and harvesting the learning from neither side.

The fix is smaller than the problem

Here is the part that tends to surprise: this is not solved with a new tool, or a project, or consulting.

A weekend solves both — with a clear rule, everyone using what works and nobody guessing where the line is.

A written rule is not a brake. It is the permission that was missing: it says what is allowed, and in saying so it frees the half that was standing still, waiting. And in saying what is not allowed, it takes off each person's shoulders a decision that was never theirs to make.

How to tell which half your company is in

Ask someone on the team, unannounced: what may you do with AI here?

If the answer comes out hesitant — or if two people give two different answers — both halves are there. And the second one, the invisible half, is probably larger than you think.