A template encodes the average business. Nothing about a real phone is average.
The failure is not that a templated line sounds bad. It is that it fails invisibly, on exactly the calls that mattered.
A templated AI receptionist is built around an average business: an average caller asking an average question with an average urgency. That business does not exist. Every business phone has a hardest decision, the call it must not get wrong, and that decision is different for a veterinary clinic than for a law firm than for a landscaper. A template cannot encode a decision it was not told about, so it does the only thing it can: it handles the call as though it were ordinary.
Why is templated failure so hard to notice?
Because it does not look like failure. The line answers, takes a name, offers a slot, and logs a completed call. Every metric a business watches says the system worked. What it did with the call that did not fit the template is not in any of those metrics.
Compare that to a phone nobody answers. That failure is at least legible; there is an unanswered call in the log, and someone can go and count them. A templated line converts a visible failure into an invisible one, which is a worse position to be in even when it happens less often.
The specific calls that fall outside a template are also, reliably, the expensive ones. A caller with an ordinary booking request is exactly who the template was built for. A caller describing a complication after a treatment, or a tenant with water coming through a ceiling, or somebody whose limitation period expires next week, is not, and those are the calls where getting it wrong costs something that cannot be recovered by calling back.
This is why the honest measure of an AI receptionist is not its answer rate or its booking rate. It is what it does with the calls it was not designed for, and no dashboard reports that.
What does a template actually get wrong?
The hardest decision on the call, because that decision is industry-specific and a template is by definition not. Every trade has one call it must not mishandle, and the shape of that call has nothing in common across trades.
The med-spa row is the one that surprises people, because the templated behaviour is *more* helpful, not less. A caller asks whether they can have filler while on a medication, and a line built to be maximally useful answers. That is practising medicine, and the fact that it sounded competent is what makes it dangerous.
| Industry | The hardest decision | What a template does instead |
|---|---|---|
| Veterinary | Is this an emergency, and should they go elsewhere right now? | Books an appointment |
| Law firms | Capture the parties before the facts, so conflicts can be cleared | Takes the story first |
| Med-spas | Price it, but never answer whether they are a candidate | Answers helpfully, which is the problem |
| Property management | Emergency or not, against a legal standard | Takes a message |
| Trades | Never commit to a price the site visit has not seen | Quotes a range it read somewhere |
None of these is derivable from the phrase "AI receptionist". Each one has to be told to the receptionist by someone who knows the business.
Do two businesses even have the same kind of phone?
Not remotely. Our own call data covers two Calgary businesses whose lines behave in opposite ways: one is a booking queue running at high volume from a single location, the other a slow trickle of first-time callers across an entire season. A single template serving both would be wrong for both.
The med-spa's line is dominated by people trying to get into a calendar, so the design problem is throughput and constraint: quoting confidently, booking against the constraint that actually binds, and refusing clinical questions. The landscaper's line is dominated by strangers asking for quotes, so the design problem is capture: getting enough detail to call back and win the job, from someone who will not call twice.
Those two lines want different opening questions, different field ordering, different escalation rules and different definitions of a successful call. The figures behind that contrast are on the report, with the arithmetic shown.
- High-volume booking lines optimise for constraint and speed.
- Low-volume stranger lines optimise for capture and callback.
- A template has to pick one, and then it is wrong for the other.
Isn't this solved by customising the greeting and the voice?
No, and conflating the two is the most common misunderstanding in this category. Choosing a voice, a name and an opening line is skinning. It changes what the receptionist sounds like and nothing about what it decides.
The decisions that matter are all invisible on a demo call: what the receptionist refuses to answer, what makes it escalate immediately, which field it confirms and which it lets go, what it does when nobody picks up the transfer, and what it may write into your booking system versus what it merely proposes.
A product can expose some of those as settings. The reason that still falls short is that the settings enumerate the decisions the vendor anticipated, and the decision that matters for your business is frequently one nobody thought to expose, an Alberta med-spa's availability following the medical director's schedule rather than the treatment room's, for instance.
A templated line does not fail loudly - it answers, books, and logs a completed call. It fails on the calls that did not fit the template, which are reliably the expensive ones, and nothing in the reporting will tell you it happened.
The test that separates the two is boring and decisive: phone the number and ask it something your business would never want it to answer.
Start free pilotWhere this argument is weakest
This page argues a position we sell. Here is where a reasonable person would push back.
- 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.
- Templates are fine for genuinely simple lines. A business whose calls really are all one shape (a single service, one location, no regulated advice, no urgency) does not have the problem this page describes, and should not pay to solve it.
- No measurement of templated failure rates. No figure is quoted for how often templated lines mishandle out-of-template calls. The argument here runs from mechanism, which is the part that holds regardless of anyone's sample.
Questions this raises
- What is a templated AI receptionist?
- One assembled from a general-purpose configuration: a standard call flow with the business's name, hours and services filled in. It encodes an average caller rather than the specific decisions a particular business's phone has to make.
- How would I know if my line is failing this way?
- You mostly would not, from reporting. The practical test is to phone it and ask the questions your business would never want answered (a clinical question, an urgency, a price on unseen work) and listen to what it does.
- Isn't a customised voice and greeting enough?
- That is skinning. It changes what the line sounds like and nothing about what it decides, and the decisions are where the cost sits.
- Are templates ever the right answer?
- For a genuinely simple line (one service, one location, no regulated advice, no urgent calls), yes. The argument on this page does not apply to a business whose calls really are all the same shape.
- Doesn't a configurable product solve this?
- Partly. Settings cover the decisions the vendor anticipated. The decision that matters for a specific business is often one nobody thought to expose as a setting.
Sources
- Recording of Customer Telephone Calls: guidance for organizationsOffice of the Privacy Commissioner of Canadaupdated 2018-04-18, consulted 2026-08-09
- Health Information Act, RSA 2000, c H-5: s. 66, information managersAlberta King's Printer / CanLIIconsulted 2026-08-09
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