Shadow AI Use: Your Staff Has Decided AI Is Worth Their Time & Their Own Money

Oct 6

Nearly a third of British workers who use generative AI do so without their employer knowing. One in six pays for at least one tool out of pocket. Together they spend close to £1 billion a year of their own money on AI for work.

Those figures come from a Deloitte UK survey of 25,000 workers, which the firm calls "the largest study of workplace GenAI use conducted in a single country." The industry has a name for this behavior: shadow AI. If you lead a nonprofit, there is a good chance it is already happening in your office.

The Secret AI Budget

So what does a billion pounds buy? Mostly email and search. Forty-three percent of users turn to AI to find information, 43% to draft emails, and 31% to summarize. No automated co-workers or assistants: just people paying their own way through an overflowing inbox.

That detail matters. Staff are not sneaking AI in to cause trouble. They use it because it saves time on the work they already do, and often because nobody has offered them an approved tool. Shadow AI is, first of all, a demand signal.

Why Nonprofits Should Pay Attention

Mundane tasks still touch sensitive data. A summarized case file, a drafted donor thank-you, a quick search on a grant applicant: each can carry names, health details, or giving histories into a tool your organization never vetted, under terms no one on staff has read.

The sector is also a prime target. Okta's Nonprofits at Work 2026 report found that nearly four in five login attempts at nonprofits are now fraudulent. Among the nonprofits Okta studied, 76% lack a formal AI strategy and 58% place no restrictions on which AI tools staff use.

Put those findings side by side and the picture is clear: eager staff, valuable data, and no fence around either.

Four Moves for Nonprofit Leaders

Governance does not have to start with a forty-page policy. These four steps bring shadow AI into the light without dampening the enthusiasm behind it.

Ask before you ban. Run a short, no-blame survey: which tools are people using, for what, and are they paying for them? You cannot govern what you cannot see.
Draw a bright line around sensitive data. Decide which information never goes into a public AI tool, such as client records, donor details, health information, and anything under a confidentiality agreement. Write that list in plain language.
Write a one-page acceptable-use policy. Name the approved tools, the off-limits data, and the person to ask when something is unclear. One page people read beats twenty pages they skip.
Pay for the tools people need. If staff are spending their own money, the demand is proven. Organization-owned accounts give you data protections and visibility, and they give your team permission to stop hiding its best habits.

Your Team Has Already Voted

Your staff has decided that AI is worth their time and, in some cases, their own money. The leadership question is whether that energy flows through channels you have chosen.

Lansby helps nonprofits write AI policies their teams will follow. If shadow AI sounds familiar, let's talk.