Using AI tools without putting customer data at risk

You paste a customer’s complaint email into a chat tool and ask it to draft a calm reply. It works well, so you do it again the next day with a quote that has their address and phone number in it. Somewhere along the way, a stranger’s personal details have left your business, and you may not know where they’ve gone or how long they’ll be kept.

None of that means you should stop using AI tools. They’re good at first drafts, summaries and tidying up wording, and we’ve written about everyday uses for admin. It does mean knowing what the tool does with your text, and keeping a few habits.

This is general information, not legal advice. If you handle sensitive data, such as health records or anything about children, speak to someone qualified.

Free tools and business plans treat your text differently

The terms behind a free consumer account and a paid business plan are often quite different, even when the tool looks the same.

On many free and personal plans, what you type can be kept and used to train the provider’s future models unless you switch that off in the settings. Some keep conversations for a set period for safety checks, and some let staff review a sample of them.

Business and team plans usually come with stricter terms. Several of the large providers state that, by default, they don’t train their models on prompts from their business products or their API, and they offer a data processing agreement. The details vary by provider and by plan, and they change.

Before anyone in the business uses a tool for work, check:

• the data or privacy settings in the account itself, including any training toggle;
• the provider’s terms for that specific plan, not the free one;
• whether it offers a data processing agreement you can accept;
• how long conversations are kept, and whether you can delete them.

If the answers are unclear, treat the tool as public and keep personal data out of it.

Leave personal data out unless the task needs it

Most tasks don’t need the customer’s name at all. Asking for a polite reply to a late delivery complaint works just as well with “the customer” in place of the name, and no order number or address.

A quick habit before you paste anything: strip out names, contact details, account numbers and anything about someone’s health or finances. If the task can’t be done without them, that’s a sign it belongs in a tool your business has a proper agreement with, or that a person should do it.

The ICO is clear that personal data processed by an AI system is still personal data, even when nobody meant it to end up there.

UK GDPR still applies to you

When you put a customer’s details into an AI tool for a business purpose, you’re the one responsible for that decision under UK GDPR. The provider doesn’t take that on for you. You need a lawful reason to use the data that way and should share no more than the task needs. If a customer asks where their details went, you should be able to tell them.

In practice that means a short written note of which AI tools the business uses, what they’re used for, and who’s allowed to use them. It also means your privacy policy should mention AI tools if customer data goes into them (our guide to cookies and GDPR for small business websites covers what a policy needs). The ICO publishes guidance on AI and data protection, which is the place to check before you do anything more involved.

Tell people if they’re talking to an AI

If your website has an AI chat assistant, say so where it starts, in plain words. People are entitled to know how their information is handled, and a chat window that looks like a member of staff can mislead them. Give them an obvious way to reach a person, and cover the assistant in your privacy policy, including who provides it and what happens to the conversations.

We build two AI agents. Bright Desk answers customer questions on a website and hands over to a person when it can’t help. Bright Ops works for the owner, drafting replies and sorting enquiries, and nothing is sent without them. Both run only on the business’s own information. There’s more on our AI agents page.

Keep a person in charge of decisions about people

An AI draft of a reply is fine. An AI deciding on its own whether someone gets a refund or a job interview is a different matter.

UK data protection law gives people safeguards when a decision with a significant effect on them is made purely by automated means, including the right to have a human look at it again. The simplest way to stay on the right side of that is to keep a person making the final call. Let the tool sort the enquiries and write the draft, then have someone read it and decide.

If you’d like a hand working out which tools suit how your business runs, tell us what you’re using now.