AI Adoption Is Not a Participation Metric

There is a slightly ridiculous thing happening inside companies right now.

Everyone wants employees to use AI.

So they buy licences. Run training sessions. Launch Copilot. Create an AI working group. Maybe even organise an AI Day.

Then six months later somebody asks:

“So... is it working?”

And nobody really knows.

They know how many licences they have.

They might know how many people logged in.

They can probably tell you how many AI experiments are running.

But none of those things tell you whether the organisation has actually got better at anything.

That is the problem.

We are measuring AI adoption like participation in the company Christmas raffle.

Using AI is incredibly easy

This is what makes the problem interesting.

AI adoption looks fast because using AI is now almost frictionless.

Open ChatGPT. Ask a question. Summarise a document. Rewrite an email. Generate a presentation.

Done.

According to Gallup's workplace research, organisations are seeing rapid increases in employee AI use. Microsoft’s Work Trend research tells a similar story: AI is spreading quickly through everyday knowledge work.

But there is an enormous difference between using AI and changing how work gets done.

An employee asking ChatGPT to make an email sound friendlier is AI usage.

A communications team redesigning research, drafting, review, approval and measurement around AI is organisational change.

Those are not the same thing.

And yet we keep putting them in the same bucket.

Check it. 👇

The employee who uses AI twenty times a day may actually be creating more work.

More drafts.

More things to check.

More mediocre content.

More information travelling around the organisation.

More output does not automatically mean more productivity.

Sometimes it just means more output.

The eternal experiment

I see another version of this when organisations talk about their AI programmes.

There are pilots everywhere.

Marketing is testing something.

HR has another tool.

Someone in finance built an agent.

The communications team has three ChatGPT accounts and a mysterious subscription nobody remembers approving.

Everyone is experimenting.

Nobody wants to stop experimenting because experimentation sounds progressive.

But eventually experimentation becomes a hiding place.

You avoid the harder questions.

What are we actually trying to improve?

Which workflows should change?

Which ones shouldn't?

Where is AI genuinely better?

Where do humans still add disproportionate value?

And what result would make us say: this worked?

If your AI strategy depends on everyone looking busy with AI, it is not a strategy. It is a participation drive with a software budget.

That distinction matters more as AI gets cheaper.

Because organisations can now generate almost unlimited amounts of work.

The scarce thing is no longer the ability to produce.

It is knowing what deserves to be produced in the first place.

Measure the work, not the AI

So I would stop asking:

“How many people are using AI?”

Start with the work.

Did the proposal take three hours instead of three days?

Did campaign performance improve?

Did the analyst find something they would otherwise have missed?

Did customer response times fall?

Did the communications team spend less time producing PowerPoints and more time thinking?

Did quality improve?

Did rework fall?

Those are much harder questions.

They are also much more useful.

Because successful AI adoption should eventually become quite boring.

Nobody celebrates Excel adoption anymore.

Nobody asks what percentage of employees are “using the internet”.

The technology disappears into the work.

That is probably where AI is heading too.

And the organisations that get there first won't necessarily be the ones using the most AI.

They'll be the ones that finally stopped counting.