Guide · AI tools
Shadow AI in small firms: how to find out which AI tools your team is really using
Shadow AI is AI your team uses without the firm knowing. Here's how UK small firms can find it, why it's a GDPR risk, and how to fix it without banning AI.
By Paulo De Sousa, Co-founder of SimpleKPI5 min read

Shadow AI is any AI tool your team uses for work that the firm hasn't approved, or doesn't know about. In most small firms it isn't sneaky. It's people trying to save time with ChatGPT on a personal login, because it works and nobody said not to.
The fix isn't a ban. It's finding out what's really in use, putting the right plans in place, and keeping client details out of the tools. This guide shows you how, in plain English.
What shadow AI looks like in a real firm
It rarely looks like a problem from the inside. Here are three made-up examples, not real clients, of the kind of thing that happens every week:
- The payroll formula. A bookkeeper can't get a formula to work, so they paste the payroll spreadsheet into a free AI account and ask it to fix it. The formula gets fixed. Every employee's name, National Insurance number and pay has now gone to a provider your firm has no contract with.
- The letter that needed tidying. A trainee solicitor uses their own Claude account to tidy up a letter to a client. It reads much better. The client's name, address and the details of their matter went with it.
- The note-taker from LinkedIn. An office manager sees a post about a new AI note-taker and signs up to try it in a client meeting. Nobody else knows the tool exists, or where the recording and transcript are kept.
None of these people did anything malicious. They were doing their jobs faster. That's exactly why shadow AI is so common, and why telling people off doesn't fix it.
Why it matters under UK GDPR
Using AI at work is allowed. The trouble starts when personal data goes into a tool your firm hasn't checked. With a free or personal account, there are four problems.
There's no contract. Your firm has no agreement with the provider, and no data processing agreement saying what they can do with what's typed. Under UK GDPR, you're meant to have one with anyone handling personal data for you (Article 28).
It may be used for training. Many free and personal plans can use what's typed to improve the provider's AI, unless someone has switched that off. You can't take it back afterwards.
There's no record. If a client, your insurer or the Information Commission (ICO) asks what was shared and where, nobody can answer.
It can be a breach. If client personal data ends up with a provider that shouldn't have it, that can be a personal data breach. Every breach has to be recorded, and the serious ones reported to the ICO within 72 hours. We cover this in more detail in can staff put client data into ChatGPT?, and our free breach deadline calculator works out the deadline for you.
Why banning AI doesn't work
The first instinct is often to ban it. It rarely works.
When firms block AI on work computers, people don't stop using it. They move it onto their phones and home laptops, because the time it saves is too good to give up. Our customers told us this again and again while we were building Saiela.
The result is worse than before. The same client details still go into the same free tools, but now the firm can't see it at all. A ban swaps a problem you can manage for one you can't see.
How to find shadow AI in your firm this week
You don't need special software to make a start. Anyone can do these four steps:
- Ask the team, without blame. Ask which AI tools they use, what for, and whether it's a work or personal account. Make it clear nobody is in trouble. You'll get honest answers only if people feel safe giving them.
- Check the company cards. Look for AI subscriptions on expenses and company cards. Paid plans are a good sign, but check they're business plans, not personal ones bought with a work card.
- Ask your IT provider. If you have device management such as Intune or Jamf, it can show which AI desktop apps are installed on work computers.
- Write it down. Put the answers into a simple AI tool register: the tool, the plan, who uses it, what for, and whether client data goes in. A spreadsheet is fine. You'll find more free help in our resources.
How to bring it into the open
Once you know what's being used, you can make it safe without taking it away.
- Pay for business plans of the tools people actually use. If half the team loves ChatGPT, a business plan with a contract is far cheaper than a breach. We compare the plans in free vs business AI plans.
- Publish a short AI policy. Say which tools are approved, what must never go in, and who to ask. One page that people actually read beats ten pages they don't.
- Allow personal accounts only for work with no personal data. Asking a free AI tool to explain a tax rule is fine. Pasting in a client's details isn't.
How Saiela helps
Saiela is a browser extension and console that does most of this for you, and keeps doing it after this week.
- See every AI tool your team uses in the browser, including new ones as they turn up. If someone starts using a new AI tool, you'll know.
- See each tool's facts by plan: whether there's a contract with your firm, whether it trains on what's typed, and where the data goes. Then approve it for client work, allow it for work with no personal data, put it under review, or block it.
- Warn people on personal accounts, or block them. They see the warning before they send, not after.
- Swap client details out of prompts, so names, emails and bank details never reach the AI, even on approved tools.
- See use by tool and by team in Activity. It counts what happened, never what anyone wrote.
AI desktop apps sit outside the browser, so Saiela reports those through device management such as Intune or Jamf instead.
Not sure where you stand? Our free AI risk check takes about three minutes and gives you a prioritised list of what to do next.
We've also written about what this looks like in your line of work, for accountants, solicitors, estate agents and small businesses.
Sources
- Information Commission (ICO): Personal data breaches: a guide
- Information Commission (ICO): Guidance on AI and data protection
- UK GDPR: Article 28, processors and contracts and Article 33, reporting a breach
Questions
The short answers.
What is shadow AI?
Shadow AI is any AI tool used for work without the firm knowing about it or approving it. Most of the time it's a free or personal account, such as someone's own ChatGPT login.
Is shadow AI illegal?
Using AI isn't. The risk is putting personal data into tools that have no contract with your firm and no safeguards, which can breach UK GDPR.
How do I know if my team uses shadow AI?
Ask them, without blame. Then check what's being paid for on company cards, ask your IT provider what device management shows, or use a tool like Saiela that shows AI use in the browser as it happens.
Should we just ban AI?
No. Bans tend to push AI onto phones and home laptops, where you can't see it at all. Approve the right tools instead, and protect the data that goes into them.
General information, not legal advice.

