Seventeen decisions sit behind an AI receptionist. A template makes all of them for you, silently.

Here they all are, so you can make them deliberately: with us, with someone else, or on your own.

Every AI receptionist embodies the same set of decisions, whether or not anyone made them deliberately. A template makes all of them by default, identically for every business using it. A built line makes them explicitly, and writes them down. The list below is the actual set; if a vendor cannot tell you their answer to each one for your business, the answers are defaults.

Decisions about how the call opens

Four, and they are the ones every caller experiences. They also carry the only clear regulatory obligation on the list.

  • 1. What the receptionist says it is. Whether it names itself as an automated system, and how early.
  • 2. The recording notice. Whether the call is recorded, stated with its purpose. This one is not optional for a business.
  • 3. How much it says before listening. Long openings increase the chance of being interrupted and the cost of each interruption.
  • 4. Whether it can be interrupted at all. Barge-in is the difference between a conversation and a menu.

Decisions about what it captures

Five, and their ordering matters as much as their content. Order by what survives a dropped call, not by what the business wants to know first.

  • 5. Field order. Callback number first makes every later error on the call recoverable.
  • 6. Which fields are confirmed. Read-backs cost seconds and patience; spend them where being wrong breaks something.
  • 7. Question shape. Closed questions survive background noise; open ones produce better detail in quiet rooms.
  • 8. Retry limits. How many attempts at a field before routing to a person. This is the setting that decides how the receptionist treats callers it understands least well.
  • 9. What it never collects. The cheapest information to protect is the kind you never captured, decisive under health privacy law.

Decisions about judgement

Five, and this is the group a template cannot make for you, because each depends on facts about your business that a category label does not carry.

  • 10. Refusal boundaries. What it answers confidently, and what it captures and routes instead.
  • 11. Escalation triggers. Deterministic conditions that fetch a human immediately, biased toward false alarms.
  • 12. Escalation fallback. What happens when the escalation destination does not answer. Most commonly missing.
  • 13. Availability logic. What actually constrains a booking, which is often not the calendar.
  • 14. Transfer mode. Warm, cold, or take a message, and which applies to which call.

Decisions about what happens after

Three, and they are the ones that get deferred until something goes wrong. Two of them are legal obligations rather than preferences.

Fifteen through seventeen are the ones businesses discover late. The first is discovered when a calendar fills with appointments nobody can honour; the second and third when someone asks a question the business has no process to answer.

  • 15. Write permissions. What the receptionist commits to your systems versus what it proposes for a human.
  • 16. Retention. How long audio, transcripts and structured data are kept, separately, because they carry different risk.
  • 17. Deletion and access. What happens when a caller asks for their recording, or asks you to delete it.

Every AI receptionist embodies these seventeen decisions whether or not anyone made them - a template makes all of them by default, identically for every business using it, and never tells you which ones it chose.

Nick Lovett, Founder, AnswerAI

Take this list to any vendor, including us. The useful signal is how many they can answer specifically for your business.

Start free pilot

How to use this against us

A checklist written by a vendor deserves scepticism. Use it symmetrically.

  1. We sell the thing this page argues for. AnswerAI builds custom lines rather than selling a configurable product, so this cluster describes our own commercial position. The arguments are made from mechanism and you should check them against a vendor who disagrees.
  2. The list favours our sales conversation. It is long, and length implies difficulty, which implies hiring someone. A business capable of answering all seventeen itself does not need us for that part.
  3. Not every decision matters equally everywhere. A restaurant does not need escalation triggers the way a veterinary clinic does. Weight the list against your own worst call.

Questions this raises

What should I ask an AI receptionist vendor?
Work through the seventeen decisions above and ask for their answer to each one for your business specifically. Vague answers on the judgement group are the signal that matters.
Which decisions are legally required rather than optional?
The recording notice, retention, and the ability to act on an access or deletion request. Those follow from privacy law rather than preference.
Which one is most often missing?
The escalation fallback: what happens when the person the call escalates to does not answer. Without it, an escalation can lose a caller more thoroughly than no escalation would have.
Can I make these decisions myself on an existing product?
Some of them, depending on what the product exposes. The list is useful precisely as a way of finding out which ones you cannot reach.
How long does deciding all this take?
With a business that knows its own phone, a couple of hours of real conversation. The engineering is not the slow part.

Sources

  1. Recording of Customer Telephone Calls: guidance for organizationsOffice of the Privacy Commissioner of Canadaupdated 2018-04-18, consulted 2026-08-09
  2. PIPEDA fair information principlesOffice of the Privacy Commissioner of Canadaconsulted 2026-08-09
  3. Health Information Act, RSA 2000, c H-5: s. 66, information managersAlberta King's Printer / CanLIIconsulted 2026-08-09
Nick Lovett

Nick Lovett

Founder, AnswerAI

Nick Lovett builds AI receptionists for service businesses across North America, and writes these from the call data they produce. Lovett Ventures Inc., Calgary.

The only real test is calling one yourself.

We build a working line on your calls, your pricing, your booking rules and your escalation rules, then you phone it and tell us what does not sound like you. About 14 days. No call limit during the pilot.

Start free pilot