What shadow AI says about leadership

A large share of people using AI at work are doing it out of their employer’s sight. Deloitte UK surveyed 25,000 UK workers for its GenAI Workforce Survey, published on 16 September 2026. Among people using generative AI, 31% said they use it without their employer knowing. On top of that, workers are spending an estimated £958 million a year of their own money on AI tools for work.

People are paying privately to do the job, and then keeping quiet about the tool. That is a management fact, not a curiosity.

Two accounts of the same office

Search work turns up both sides of this in the same week. Candidates describe a workflow in which half the day runs on a tool the company has never approved. A hiring manager, speaking about the same kind of team, will say the group is not really using AI yet. Both accounts can be true. The manager is describing the official stack. The candidate is describing the work as it is done.

The number that should concern a board more than the hidden tools themselves comes from the same survey. 65% of GenAI users say there is a lack of convincing leadership on how AI should be used at their organisation. Where the rules are vague, people invent their own. Some of those private choices will be harmless. Some will send customer data, source code or board papers to a product nobody has checked.

What a ban does

A blanket ban feels decisive. It also pushes the same behaviour further under the desk, because the work that the tool was doing does not disappear. Staff who already pay for a product out of their own pocket have shown they will keep going if the approved route is slow or missing. The safer response is a short list of tools, a plain rule on what data may leave the building, and a named owner when something goes wrong.

That owner is often missing. AI use has spread across departments, each of which bought or subscribed on its own, and nobody can say what it costs in total or whether the data is safe. That is one of the signs that a firm may need a Chief AI Officer, a Head of AI, or a clearer remit for the CTO. The title matters less than the accountability. Someone has to be able to stop a tool, explain the risk to the board, and give staff a route that is easier than the unofficial one.

What to ask before a senior hire

A technology leader walking into this will ask three practical questions in the first conversation.

Which tools are people actually using, including the ones they pay for themselves? Who is allowed to approve a new one? What happens to customer and company data if a tool is wrong, or if the supplier changes its terms?

If the honest answer to the first question is “we have not asked”, the brief for a Chief AI Officer or a CTO should say so. Pretending the estate is tidy, when staff already know that it is not, is how a senior hire loses trust in the first month. Clario puts that in the written brief, with the reporting line and the salary, before anyone is approached.

The survey figures above are from the Deloitte UK GenAI Workforce Survey of 16 September 2026, cited here as published research. A board can start with one question to the people who do the work: which AI tools are you using, including the ones the company has never approved?

The Chief AI Officer page sets out when that question needs its own executive, and when it still sits properly with the CTO. The guide to hiring a first Chief AI Officer covers the brief, the interviews and the first 90 days.