Should your AI sales assistant say it's an AI?

Some AI sales tools sell their ability to pass as human: a made-up name, a photo, a reply that insists it is a real person. It is tempting. It is also a bad idea, for reasons that have little to do with ethics and a lot to do with keeping customers.

People find out

A customer who discovers that the friendly Sarah who emailed them twice does not exist does not think your business is clever. They wonder what else is not real, starting with your prices and your reviews. Trust lost that way is very hard to win back, and it is usually shared.

The law is moving one way

A growing number of rules require that people are told when they are dealing with an AI. California's bot disclosure law covers bots used to encourage a sale, and the EU AI Act sets out transparency duties for AI systems that interact with people. The detail differs from place to place, but the direction does not: being clear about what is talking is becoming the expectation.

Disclosure costs less than you think

What makes someone reply to a sales email is relevance: a specific reason you are writing to them, from a real business they could look up. Whether a person or software typed the first draft matters much less than whether the message is useful. A message written for a named business, signed by its owner and sent from the owner's own address is honest about what it is and still personal.

What disclosure looks like in practice

Disclosure does not mean opening every email with a warning. It means never pretending, and answering straight when asked. In practice:

• The email is from your business, sent from your address, and says who it is from. No fictional colleague signs it.
• When someone asks whether a person wrote it, the reply says an assistant drafted it for the business, and that the owner reads every reply.
• In a chat, the assistant says it is the business's assistant when asked, and offers to bring in a person.
• Nothing is claimed that the assistant cannot back: no I remember you from the conference, no fake personal anecdotes.

When a person should take over

Some conversations should not be handled by software at all, however good it is. A customer who is upset, a complaint, a request for a refund or an exception, anything involving health, money owed or legal questions: these should reach you quickly, with the history attached, so the customer does not have to explain twice. An assistant that knows when to step back protects your business far more than one that tries to handle everything.

A simple policy

1. Send as your business, from your own address, under your own name.
2. Never invent a person. No fictional sales rep, no stock photo.
3. Say yes when asked. If someone asks whether they are talking to an AI, the answer is yes, and who it works for.
4. Hand over to a person when someone asks for one, or is unhappy.

Questions to ask before you buy a tool

• Can it be set to impersonate a person? If the answer is yes, ask who else using it has turned that on, and how it reflects on the whole category.
• What does it say when asked? Try it. Ask the demo directly whether it is a bot.
• Does it hand over? There should be a clear route to a person, with the conversation attached.
• What does it send without you? The safest answer, at least at first, is nothing.

How Speek handles it

Speek says yes when asked, never claims to be a named person, and says which business it is writing for. You choose how it sounds; you cannot make it pretend to be human. Everything it sends to a customer waits for your approval first.

This is general information, not legal advice. Check the rules where you and your customers are.