What happens when an AI receptionist doesn’t know the answer?

Give it the answers you actually want it to use

The first thing I’d ask for is the information your business already knows: services, hours, service area, booking rules, and answers to the questions you get all the time. If you publish a standard service-call fee, that can be an approved answer. If you quote every job individually, that needs to be clear too.

That information is the starting point, not a guarantee that an AI system will never make a mistake. I’d keep its responsibilities narrow, test the questions that could cause problems, and make it easy to get a person involved.

When your hours, prices, or policies change, the receptionist’s information needs to change with them. Keeping that current belongs in the agreed care arrangement.

An honest answer is better than a confident guess

Imagine a caller asks how much it will cost to replace a specific AC part. If your business hasn’t approved a price for that situation, a considerate AI receptionist should say the technician needs to assess it. It can then offer the next allowed action, such as booking a diagnostic visit or taking the details for your team.

That still helps the customer. They know why they aren’t getting a price immediately, and they have a way forward. It also keeps the receptionist from making a promise you’ll have to undo later.

• Use the prices and policies you have approved.
• Acknowledge when the question needs your judgment.
• Offer a booking, message, or transfer that fits that situation.
• Never describe an unsuccessful transfer as a completed conversation.

“Can I speak to a person?” needs a real plan

If someone wants a person, I wouldn’t make them argue with the receptionist. We’d decide who gets the call, when that person is available, and what happens if they don’t answer.

For your Arcadia Websites setup, I configure call transfers to your chosen person or team’s phone number. Transferring a caller to another AI agent isn’t the same as connecting them with a person.

There’s an important difference between having a transfer option and having live staff available around the clock. Arcadia Websites doesn’t include a 24/7 human call center. If nobody can answer, the fallback might be a clear message, a callback request, and a notification to the person responsible.

Test the awkward calls before customers make them

The easy demonstration is someone asking your opening hours. The useful test is someone asking for an exception, changing the subject, or insisting on a person after hours.

I’d walk through those situations with you before launch, including an unknown price, an unavailable transfer destination, and a booking that can’t be completed. We’d check the spoken response and what reaches your team afterward.

No AI tool is mistake-proof. The aim is to give a professional AI receptionist a clear job, limit what it can promise, and make sure there’s a sensible route when it reaches that limit. You should understand that route before the first customer calls.