Should your AI receptionist disclose that it's an AI? (Yes.)

Agentric discloses that callers are speaking with an AI at the start of every call. Not buried in a disclosure — plainly, upfront, before the conversation starts. Here's why that's non-negotiable, and why it works better anyway.

The legal picture

UK GDPR doesn't contain a clause saying "disclose your AI" — but transparency is a core principle, and ICO guidance on AI and data protection points firmly toward telling people when they're interacting with automated systems. Consumer protection law adds another layer: the Consumer Protection from Unfair Trading Regulations ban misleading practices, and an AI performing human-ness to callers arguably qualifies.

The direction of travel is unambiguous — the EU AI Act explicitly requires AI transparency for natural-language systems, and UK regulators are rowing the same way. Disclosing now is both compliant now and future-proof.

The trust argument

Here's the counterintuitive part: disclosure helps the calls. When the assistant says "you're speaking with our AI assistant":

• Expectations calibrate — callers stop testing for humanity and start asking for what they need.
• Anything impressive is a pleasant surprise; nothing feels like deception.
• The minority who want a human say so immediately → faster handoff → less frustration.

Businesses that tried the "hide it" approach found the same thing repeatedly: callers who suspected and confirmed AI felt deceived; callers told upfront mostly just… got on with it.

How to disclose well

The formula is one clause, matter-of-fact, benefit-adjacent:

"Thanks for calling Hartley Heating — you're speaking with our AI assistant, and this call may be recorded. How can I help?"

What makes it work: it's brief, it doesn't apologise, it pairs the disclosure with a capability promise (recording notice + immediate "how can I help").

The disclosure should also appear in the chat widget ("AI assistant — not a human") and on your landing page footer. Consistent transparency reads as professionalism, not caveat.

The one rule

You may soften or shorten the disclosure in settings — but the assistant will never claim to be human if asked, and that's a hard constraint. "Are you a real person?" always gets an honest answer.