Do I have to tell customers I'm using AI?

You are using AI somewhere in the business now, drafting, replying, maybe a chatbot, and a small voice asks whether you are supposed to be declaring it, like an allergen. The honest answer has a legal part, which is narrower than people fear, and a trust part, which is where the real decisions live.

The legal part

UK GDPR does not require a general announcement that your business uses AI. What it regulates is solely automated decision-making with legal or similarly significant effects: a machine deciding something consequential about a person with no human involved. Refusing credit, filtering job applicants out, setting an insurance price. If you do that, you must tell people it is happening, explain the logic in meaningful terms, and give them a route to a human review. The ICO's guidance on AI and data protection covers the detail, and our AI and GDPR guide translates it for small firms.

Most small-business AI is nowhere near that line. AI drafting an email a human reads and sends is not an automated decision; the human is the decision. Transcribing meetings, categorising your books, writing first drafts of quotes: none of it triggers a disclosure duty. Where you should pause: anything that auto-rejects people (job applications, tenancy screening, bookings filtered by risk) without a human look. That is the territory where the law has opinions, and "the software decided" is precisely the thing you must own and explain. Recruitment has its own tripwires, covered in using AI in recruitment without breaking UK law.

The trust part

The better question than "must I disclose" is "would the customer feel deceived if they found out". That test sorts every case I have met.

A chatbot that opens with "Hi, I'm Sarah" and turns out to be software fails it instantly. Nobody minds talking to a bot that says it is a bot and solves the problem; people mind being fooled, and they remember it. So the rule for anything conversational is: label it, give it a way to reach a human, and let it be useful instead of pretending. Our answer on whether customers can tell they are talking to AI goes deeper.

AI behind the scenes passes the same test without any announcement. Your customers do not expect a declaration that you used spellcheck, a template or an assistant to draft the email you chose to send. What they are owed is that the words are true and that you stand behind them, which has nothing to do with who typed the first draft, and everything to do with who checked it.

In my own work the line is simple, and I commend it to you: I never hide it, I never lead with it. Customers are buying the result. Be straight when software is the thing they are talking to, take responsibility for everything that goes out, and treat AI like any other tool in the business: yours to use, and yours to answer for. Do that, and disclosure mostly takes care of itself, one honest label on the chatbot and a clear conscience everywhere else.