The permission gap

Ask why AI adoption inside most organisations is so patchy and skill turns out to have little to do with it. Most people can work out how to draft a letter or summarise a long report with one of these tools in an afternoon. They hold back because nobody with the authority to say so has ever said, out loud, that using the tools is allowed.

Without that, people split, often inside the same team. Some don’t touch the tools at all. They’ve heard enough about data ending up somewhere it shouldn’t, or a wrong answer delivered with total confidence, to decide that silence means no. Others use the tools anyway, on their own devices and in their own time, because the work is quicker and they haven’t been stopped. Neither group has been given anything to go on, not even somebody to ask.

The cost of silence

A leader who has said nothing has still made a decision without noticing it. Silence reads as caution to somebody who is already cautious and as a green light to somebody who is already confident. The organisation ends up governed by whichever temperament happens to be sitting at each desk.

Most leaders don’t see it that way. They notice their teams aren’t using the tools much, decide it’s a training gap and book a session. But training won’t help until someone has said the tools are allowed and what sensible use looks like, and only whoever holds the authority can say that.

I understand why leaders hesitate. Taking a position means admitting how much you don’t yet know about a tool that changes every few months, and somebody might later point at what you wrote and show you got it wrong. Saying nothing avoids all that. It also moves the cost onto everyone below you, who can’t tell whether they’re being careless or sensible because the line hasn’t been drawn.

Policies that do arrive tend to rest on assumptions. They assume people understand the risks and will ask if they’re unsure. They assume the existing rules about confidentiality cover a tool that didn’t exist when those rules were written. It would be better to ask people what they’re already doing and where they’re getting stuck. Then ask what would need to be true for them to feel safe using the tools on real work. That conversation can happen well before any document does.

Some of the caution is well placed. A lot of the work these tools get pointed at touches information people were told, or assumed, would be handled carefully. It might be someone’s personal circumstances, a complaint, notes from a difficult conversation or records about somebody who isn’t there to consent to how their information gets used. The care that information is owed doesn’t go away because a member of staff finds a tool useful. Say nothing, and each person makes that judgement alone, case by case, without anyone more senior having thought it through first.

Saying it out loud

The statement itself can be short, and it doesn’t need a year of drafting. A page will do, saying what the tools may be used for, what they must never be used for and who to ask when a case doesn’t obviously fit either. It should be published somewhere everyone can find it, dated and owned by a named person, so there’s somebody to go back to when the tools change again.

It won’t be complete, and it won’t survive the next version of the tool unchanged. It can be revised when that happens, and people can use it in the meantime.

Two small organisations show how little a statement needs. Crowborough Town Council, in East Sussex and not far from where I live, adopted a two-page artificial intelligence policy in March 2026, listed on its policies and governance page.

It names ChatGPT, Copilot and Claude directly. Drafting reports, minutes and publicity is allowed with human review. Decisions with legal effect on individuals, any form of surveillance and putting personal data into public AI tools without safeguards are not. The town clerk oversees it, and the council wrote it for itself.

Action for Happiness, a UK charity, publishes its AI policy on an ordinary page of its website, with a commitment to say where the tools helped produce published material. It’s longer than Crowborough’s and still readable in one sitting.

Who is deciding this in your organisation right now, just by staying silent, and would they have made that decision on purpose?