Telling customers they are talking to AI: a UK guide
By Keith Hamilton · · 4 min read
Telling customers they are talking to AI is simple in principle: if a customer is dealing with AI, they should know. The details are where businesses get it wrong, with a chatbot given a human name and photo, a voice agent that dodges the question, or an AI-written reply signed by someone who never saw it. This guide explains why being open matters, when to tell people, what to say, and how to make sure a person is always within reach.
Why does AI disclosure to customers matter?
AI disclosure to customers matters because it protects trust and keeps you on the right side of the law. A customer who discovers that the friendly "Sarah" on your website chat was a machine starts to wonder what else you have not told them. Telling them at the start removes that problem before it exists.
There are legal reasons too, though this is not legal advice:
- UK GDPR requires you to be open about how you use people's personal data. If AI is handling their conversations, analysing their calls or helping make decisions about them, your privacy information should say so.
- Consumer protection law prohibits misleading practices. Presenting an AI as a person, in a way that affects a customer's decisions, risks falling foul of it.
- Regulated businesses, such as those authorised by the FCA, need to check what their regulator expects on fair treatment and clear communication.
- If you serve customers in the EU, the EU AI Act includes transparency duties for AI systems that interact with people.
When should you tell customers they are talking to AI?
Tell them at the start, whenever a customer is dealing directly with AI. That one rule covers every channel:
- Website chat and messaging. Say that it is an AI assistant before the conversation begins, not in small print.
- Phone calls. The voice agent introduces itself as AI in its opening line.
- Outbound calls and messages. Say who is contacting them, why, and that it is an automated assistant.
- Automated emails. If replies are written and sent by AI with no person checking them, make that clear.
Where AI only helps a member of staff, the position is different. If a colleague uses AI to draft a reply, then reads it, corrects it and sends it as their own, that person has taken responsibility for the message.
In that case you can mention AI assistance in your privacy information rather than on every email. Whatever you decide, write it down in your AI policy so the whole team does the same.
What should an AI chatbot or voice agent say?
It should say, in one or two plain sentences, that it is AI, what it can help with and how to reach a person. The aim is to tell people what they are dealing with, not to apologise for it:
- Chat: "Hi, I'm the [business name] AI assistant. I can help with orders, deliveries and opening hours. You can ask for a member of the team at any time."
- Phone: "Thanks for calling [business name]. You're speaking to our AI assistant. I can help with most questions, or put you through to someone if you'd prefer."
Giving an assistant a name is fine, as long as it is clearly presented as AI. Avoid human photos, invented job titles or anything else designed to make people think they are talking to a person. This is the core of AI chatbot transparency in the UK.
If a customer asks "Am I talking to a real person?", the answer must be honest. Build that into the assistant's instructions and test it before launch.
How do you make a person easy to reach?
Let customers ask for a person at any point, in their own words, without being sent round in circles. Telling people they are talking to AI only helps if they can do something about it.
When nobody is available, the AI takes a message, books a callback or sends a text with useful information. When a conversation is passed on, the colleague sees what has already been said, so the customer does not have to start again. Our guide to blending AI and people in a contact centre covers hand-overs in more detail.
Take particular care with customers who show signs of vulnerability. Distress, confusion, illness or financial difficulty should lead to a person, not more automation.
What are the rules on recording and outbound AI calls?
Recorded or transcribed AI conversations follow the same personal data rules as any other call. Tell people at the start, explain why in your privacy information, and decide how long you keep recordings.
If you plan to use AI for outbound calls or messages, check the Privacy and Electronic Communications Regulations (PECR) and the ICO's guidance carefully. There are strict rules on automated marketing calls and on marketing texts and emails, and consent is often required.
Service calls, such as chasing a payment, are treated differently from marketing, but they still need to be fair and transparent.
How do you check AI voice agent disclosure works in practice?
Test the experience as a customer would before going live:
- Is it clear within the first few seconds that this is AI?
- Does it answer truthfully when asked if it is a person?
- Can you reach a person easily, by asking in different ways?
- Does the person you reach know what has already been said?
Then keep checking. Review a sample of conversations regularly, look at complaints and listen to what customers say about the experience. Our guide to AI voice agents for customer calls sets out the measures worth tracking.
Where Forwardcycle fits
In the voice agent work we have done, conversations the AI cannot resolve are passed to a person, and in our council proof of concept callers are transferred whenever they want to speak to someone. Our safe adoption service helps you set clear, honest rules for how AI talks to your customers, as part of an AI policy and staff guidance. To see where you stand, take the free AI Readiness Assessment.