AI receptionists, honestly

What an AI receptionist actually is

A voice agent that answers your phone, holds a real conversation, and does something with it — answers the question, books the appointment, triages the emergency, routes to a person.

The distinction worth holding onto is between answering and resolving. An answering service answers and takes a message. An AI receptionist is only worth buying if it resolves, because a message-taking AI has all the limitations of automation and none of the benefit.

The problem it solves, stated plainly

The missed call. For a service business a missed call is usually a job that went to whoever answered on the second attempt, and the calls most likely to be missed — nights, weekends, during another call — are disproportionately the urgent, high-value ones.

You can measure your own exposure without buying anything. Pull your call log for a month and count calls that rang out, went to voicemail without a message, or came in while another was active. Multiply by your average job value and a conservative conversion rate. That number is the honest size of the problem, and it is yours rather than a vendor's.

What a good one does on a call

• Answers immediately, on every line. No hold, no queue, no second ring.
• Identifies itself as an AI. Non-negotiable, for trust and increasingly for law.
• Answers the actual question from your knowledge — service area, what you do, rough cost, availability.
• Triages by urgency. The single most important behaviour. A flooding bathroom and a question about hours must not be handled the same way.
• Books into real availability, not a slot you do not have.
• Escalates without a fight. Anyone who wants a person gets one, quickly.
• Logs everything — transcript, outcome, details captured.

Where AI receptionists genuinely fail

If a vendor will not tell you this part, that is informative:

• Emotional distress. A frightened, grieving or furious caller needs a human. Detect and transfer; do not try to handle it.
• Heavy accents and poor connections. Recognition degrades. A good deployment escalates after two failures rather than asking a third time.
• Anything genuinely unanticipated. It follows rules. Novel situations need a person.
• Complex multi-party scheduling. Coordinating three people's availability is still hard.
• Callers who refuse to engage with automation. Some will. Get them to a human fast rather than winning the argument.

What it costs, and how to check

Two components: a one-off build fee covering the call inventory, knowledge and rules, and a monthly capacity allowance tied to call volume and length.

We are not going to publish a comparison table of competitor pricing, because those figures vary by provider and market and a fabricated table would be the least trustworthy thing on this page. What you can do instead is concrete: take your current answering service invoice, or the loaded cost of the front-desk hours spent on the phone, and your own monthly call count. We will put our numbers alongside yours on a call. If yours are lower, we will say so.

The structural difference is that per-message and per-minute pricing rises in direct proportion to how busy you are, so you pay most in your best month. Capacity-based pricing does not. The full comparison is here.

How to deploy one without annoying anybody

1. Inventory your calls first. Listen to a week. Most businesses have four or five real call types and they are not the ones on the website.
2. Write the triage rules before anything else. What is an emergency, what happens to it. Build and test this path first.
3. Start at the edges. After-hours and overflow. You get the benefit where you currently have nothing, and you build trust before it touches your main line.
4. Listen to real calls in week one. Not a sample — all of them, at first. This always changes something.
5. Make the human path obvious. The fastest way to make customers hate an AI is to make escaping it difficult.
6. Expand only once the transcripts are boring.

Which industries this works best in

Emergency-heavy, appointment-driven and intake-driven businesses, because all three have high call volume, predictable call types, and a direct link between answering and revenue. We have written up the specifics per trade:

• HVAC contractors — The no-heat call at 11pm
• Plumbing companies — Water is damaging property while the phone rings
• Electrical contractors — The safety question comes first
• Medical and dental practices — High volume, and a privacy constraint
• Law firms — Intake quality decides the matter
• Pool and outdoor services — Three businesses, one phone line