Every owner who sets up an AI front desk arrives at the same uncomfortable question, usually late in the process: do I have to tell people?
The instinct to stay quiet is understandable. You worry that "you are talking to an assistant" makes you sound smaller, or cheaper, or less personal than the competitor down the road. That instinct is worth taking seriously, and it is also mostly wrong.
What actually happens when people find out
The damage is almost never caused by the AI. It is caused by the discovery.
Somebody who is told up front that they are talking to an assistant, and who gets fast, accurate, useful answers, generally does not mind at all. Somebody who assumed they were talking to you, and works it out halfway through, revises their opinion of everything else too. They start wondering what else was not what it appeared to be.
That is the real asymmetry. Disclosure costs you a small amount of perceived personal touch. Concealment risks the one thing a sole proprietor genuinely cannot replace, which is being believed.
The rules are tightening, and they vary
Beyond trust, there is a compliance dimension that has moved quickly. Transparency obligations for systems that interact directly with people now exist in the European Union, and several US states have their own disclosure requirements, some limited to particular sectors or to answering honestly when asked.
We are not lawyers and this is not legal advice. The rules differ by where you are and what you do, and they are still changing. But the direction of travel is clear enough to plan around: the number of places where quietly passing off software as a person is acceptable is shrinking, not growing.
Designing for disclosure now is cheaper than retrofitting it after a rule change, or after a complaint.
What disclosure does not require
Being honest does not mean leading with a disclaimer. You do not need a banner announcing a robot, you do not need to interrupt a good conversation to restate it, and you do not need to make it the first thing anybody reads.
What you need is that nobody is ever misled. Those are different standards, and the second is far easier to live with.
Wording that works
The pattern that holds up is simple: name it once, early, plainly, then get on with being useful.
- Do: "Hi, I am the assistant for Doherty Plumbing. I can take your details and get Sean to call you back."
- Do not: "Hi, you are speaking with Sean." when you are not.
- Do: give it a role rather than a fake identity. An assistant is honest. A fabricated colleague called Sarah is not.
- Do: make the handover explicit. "Sean will read this himself and reply on Monday."
Notice that the honest version is not apologetic. It does not say "unfortunately you have reached an automated system." It states what it is and immediately says what it can do for you, which is the part the caller cares about.
The one place you must be unambiguous
If somebody asks directly whether they are talking to a person, the answer is always the truth, immediately and without hedging.
This is the line that matters most, and it is the one worth checking before you go live. Ask your own system, in those words, and see what it says. If the answer is evasive, fix it before a customer finds it. A system that dodges that question will eventually do it to somebody who then tells other people.
Where this leaves a business of one
The honest setup is not the weaker one. "I am a one person business, and I use an assistant so that nothing gets missed while I am on a job" is a perfectly good story. It explains why you answer at nine at night when your larger competitor does not, and it makes you sound organised rather than small.
The same principle applies when you serve customers in more than one language, which we cover in serving Spanish-speaking customers: being clear about what the system is costs nothing, and being caught out costs a great deal. Our features page shows what the assistant actually says and where you take over.
