What questions can an AI receptionist answer?

An AI receptionist answers any question your business itself can answer — hours, location, services, published pricing, availability, policies, and the status of an existing job — and it does more than answer: it books, quotes from your price list, screens, and transfers. What it will not do is guess. Anything outside what it has been trained on gets an honest "I don't know" and an escalation to a person. BotsDontSleep runs this 24/7 in 90+ languages at $0.60 CAD per answered minute, and you can test all of it on our demo line at +1 437-494-9110.

Questions about your business

The highest-volume category, and the one that eats the most front-desk time while converting nobody.

• "What time do you close on Saturday?"
• "Are you open on the holiday Monday?"
• "Where exactly are you — is there parking?"
• "Do you service Etobicoke, or just Mississauga?"
• "Do you take walk-ins?"
• "Do you do transmissions, or just brakes and tires?"
• "Do you take my insurance?"
• "Are you accepting new patients?"

These come from your own material — hours, service list, coverage area, policies — loaded during setup. A receptionist answering the parking question for the ninth time that day is the clearest case for handing a category to software.

Questions about booking

• "Can I get in Thursday afternoon?"
• "What's your earliest opening?"
• "I need to move my appointment on the 14th."
• "Can I cancel and rebook for next week?"
• "How long does the appointment take?"
• "Which of your technicians can do this?"

These are not answered, they are done. The agent reads live availability, offers real slots, writes the appointment in, and confirms by text before the call ends. Reschedules at 9pm are the ones that matter most, because the alternative is a no-show. There is more on this in will an AI receptionist book appointments.

Questions about price

• "How much is a standard cleaning?"
• "What do you charge for a consultation?"
• "Is the diagnostic fee applied to the repair?"
• "Do you charge extra for evenings?"

The agent quotes from your published price list and nothing else. If a job needs to be seen before it can be priced — most real quoting work does — it says so, captures the details, and books the estimate or the callback instead of inventing a number.

This is the boundary worth being blunt about. An agent that guesses at a price to keep a caller happy has committed your business to a figure you never approved. Ours does not do that, and if you are evaluating another one, this is the first thing to test.

Questions about an existing job or file

• "Is my car ready?"
• "Has the report been sent to my lender yet?"
• "Did you get the documents I emailed?"
• "What's the status of my claim?"

Whether these get a real answer depends on one thing: whether the agent is connected to the system holding that status. Connected, it looks it up and tells the caller. Not connected, it takes a structured message and routes it to the person who can answer, with everything the caller already said attached — so nobody repeats themselves.

Status calls are worth counting before you deploy. In most service businesses they are a much larger share of daily volume than owners expect, and they are almost entirely automatable.

Questions that reach a person

Some questions should never be answered by software, and the rules are yours to set. In our deployments the standard set is:

• Advice — medical, legal, or financial. The agent takes the intake and books the professional.
• Complaints about your work. These go to a person immediately, with the transcript.
• Negotiation. Discounts, payment plans, disputed invoices.
• Emergencies. No heat in February, water through a ceiling, a distressed pet owner. Trades and veterinary deployments treat these as transfers, not bookings.
• Anything untrained. If it is outside what it knows, it says so and escalates.

You can make that list as narrow or as wide as you want. "Transfer any caller who asks for a person" is a valid rule, and plenty of businesses use it.

Which questions should you test on a demo line?

The useful test is not whether it can answer easy questions. It is whether it fails honestly. Call +1 437-494-9110 and work through these eight — they take about five minutes together.

1. Ask something it cannot possibly know. "What was my last invoice number?" Good: it says it does not have that and offers to take a message or transfer. Bad: it produces a number.
2. Interrupt it mid-sentence. Start talking over the greeting. Good: it stops and listens. Bad: it finishes the sentence, or restarts.
3. Change your mind halfway through. Ask for Tuesday, then say "actually, Thursday." Good: it follows without resetting. Bad: it starts the booking over.
4. Switch language mid-sentence. Begin in English, finish in Punjabi, Hindi, or Tagalog. Good: it continues in the language you switched to. Bad: it restarts the call in a new language, or ignores the switch.
5. Ask a vague question. "Do you do the thing with the brakes where they squeal?" Good: it asks a clarifying question. Bad: it says it did not understand.
6. Ask whether it is an AI. Directly. Good: it says yes plainly. Bad: it deflects.
7. Ask for a person. "Can I just talk to somebody?" Good: it transfers or takes a callback, without an argument. Bad: it tries three more times to help you.
8. Ask a price question it has no rule for. Something unusual and specific. Good: it says the job needs to be looked at and offers to book the estimate. Bad: it quotes.

Run the same eight against any AI receptionist you are evaluating, ours included. Questions 1 and 8 are the ones that separate products.

Who answers what?

What should you load it with before launch?

The scope of what it can answer is set by what you give it, and the businesses that get the most out of the first month bring four things to the discovery call.

Your hours, holiday closures, location details, and coverage area. Your service list, including what you do not do — the exclusions prevent more bad bookings than the inclusions enable. Your published prices and the rules around them. And the twenty questions your front desk actually gets, which is the most valuable document and the one almost nobody has written down.

Whoever answers your phone today can list those twenty questions in about ten minutes. That list is most of the build.