10 questions to ask before choosing an AI receptionist

One: does it book appointments, or just take messages? This is the dividing line in the whole category. A system that takes a message has automated your voicemail. A system that books the job during the call — and sends the caller an SMS confirmation before they hang up — has automated your front desk. The caller who has to wait for a callback is a caller who can still ring your competitor in the meantime.

Two: does it answer from your knowledge, or from a script? Callers ask real questions — do you do this, what does it cost, are you open Saturday. A good system answers from your own information: your services, your prices, your hours, your policies. If the answer to “how does it know our prices?” is vague, keep shopping.

Three: does it work on your existing number? Your number is on your van, your website, and ten years of invoices. You should not have to print new cards or retrain your customers to adopt a piece of software. The AI should answer the number you already have.

Four: can you read and hear every call? Every call should be recorded and transcribed somewhere you can actually check — one dashboard, not a weekly summary email. In the first fortnight you’ll want to read everything; it’s how you build trust in the system and how you catch gaps in the information you’ve given it.

Five: what happens to callers who don’t book? Plenty of callers aren’t ready to commit on the spot. A good system captures them as leads — name, number, and what they wanted — so you can follow up. A caller who rang, asked, and vanished without a trace is the old problem wearing new clothes.

Six: what happens when a call needs a human? Some calls do — complaints, emergencies, callers who simply insist on a person. Ask exactly how escalation works: does it transfer, does it notify you, does it take a message? “The AI handles everything” is not a reassuring answer; it’s a red flag.

Seven: what does it do when it doesn’t know? Ask the vendor directly. The right behaviour is to say so, take the caller’s details, and flag it for you — not to improvise. An AI that guesses at your prices is worse than one that admits the gap.

Eight: are you locked into one AI provider? Under the hood, these products run on a language model and a text-to-speech voice. The field is moving fast, and today’s best model or most natural voice may not be next year’s. A well-built product lets you choose and change providers; a poorly built one welds you to whatever it launched with, and you inherit its limitations for as long as you’re a customer.

Nine: how does the pricing actually work? Understand what you’re billed on — minutes, conversations, or something murkier — and what happens in a busy month. Per-minute rates that look small on a pricing page compound quickly on a phone line that’s doing its job.

Ten: do cheaper plans get a worse product? Some vendors gate the important features — booking, transcripts, escalation — behind the expensive tier, which means the plan you can afford doesn’t solve the problem you have. Better structures give every plan the full product and vary only the capacity.

Take these questions into every demo and make the salesperson answer all ten plainly. Any product that handles them well is worth trialling; any that dodges two or more is telling you something.

For transparency, here is how Estric answers them: it books during the call and confirms by SMS, answers from your own knowledge base on your existing number, records and transcribes every call into one dashboard, captures non-bookers as leads, escalates to a human when needed, and lets you choose your AI model and voice provider. Plans are $99, $249, and $599 a month, every one of them the full product — the tiers differ only in monthly voice minutes and conversations.