Off-the-shelf AI receptionist, or a custom build?

Every software company will tell you to buy their software. Every agency will tell you to commission a build. I sell the expensive option, so treat this accordingly, but I would rather tell you the truth and lose the sale than take money for something you did not need.

Start here: most businesses should buy the cheap tool

If you are a solo operator or a small team who wants the phone answered, appointments booked into one calendar, and messages sent to you, buy an off-the-shelf AI receptionist and do not hire anybody.

The products in this space start around twenty-five to fifty dollars a month. They are genuinely good now, they set up in an afternoon, and for a straightforward business they will do roughly ninety percent of what a custom build does at a fraction of the cost. Rosie, Voksha, My AI Front Desk, Dialzara, Synthflow and several others all sit in this bracket.

If that describes you, you have your answer. Go and buy one. Genuinely.

Where off-the-shelf starts to break

The products are built for the common case, which is what makes them cheap. They start creaking when your business is not the common case. In practice that happens for five reasons.

1. Your systems do not have an integration

The tools integrate with the popular calendars and the popular CRMs. If you run a niche field-service platform, an industry-specific practice system, or something built for you years ago, there is no integration and there never will be, because you are not a big enough market.

The tell: you are being asked to change the software your whole business runs on so that your phone system can talk to it. That is the tail wagging the dog.

2. Your routing is genuinely complicated

Multiple locations with different opening hours. An on-call rota that rotates weekly. Different engineers for different job types. Escalation that depends on the customer, not just the request.

Off-the-shelf routing tends to assume one business, one calendar, one set of hours. Once you need real conditional logic, you are fighting the product.

3. The call has to trigger work elsewhere

This is the most common reason people come to me, and it is not really about the phone at all. The call is the start of a process: create the job, notify the engineer, send the pre-visit form, add them to the follow-up sequence, raise the invoice when it is done, ask for the review a week later.

A receptionist product answers the phone. It does not run your business. If what you actually need is the whole chain automated, buying a phone tool solves the smallest part of it.

4. You have compliance or data constraints

Where recordings live, how long they are kept, who can access them, what has to be logged. If you have real obligations, "the vendor handles it" is not always an acceptable answer, and you may need the pieces in accounts you control.

5. You already tried and it did not work

Worth saying, because a good share of the people I talk to are here. They bought the tool, it did eighty percent, and the missing twenty percent was the part that actually mattered. That is not a failure of the product. It is a signal that the business is past what a product can do.

The crossover, in plain arithmetic

Off-the-shelf is a low monthly cost and no build cost. Custom is a real build cost and then a lower running cost, because you are paying for the underlying services rather than a subscription with margin on top.

Purely on cost, off-the-shelf wins for a long time. A build has to be justified by something other than the monthly line, and that something is almost always one of two things:

• Work the product cannot do at all, where the comparison is not cheaper-versus-dearer but working-versus-not.
• Staff time it removes elsewhere. If answering the phone properly means someone re-keying every call into three systems, the phone tool has not saved you the expensive part.

My blunt test: if you can write down what you need in one sentence and it does not mention any system other than a calendar, buy the product. If the sentence needs an "and then" three times, that is a build.

The dishonest arguments, from both sides

What software companies overstate: the breadth of integrations, how much conditional routing you can really configure, and how far the free plan goes before the useful features start.

What agencies like mine overstate: how bad the cheap tools are. They were poor two years ago. They are not any more, and anyone still describing them as robotic phone menus is either out of date or hoping you are.

The genuinely honest framing is that the products got good enough to serve the majority of the market, which pushed custom work up into the cases that are actually complicated. That is a healthier place for both to sit.

What to do this week

1. Write the one sentence. What do you want to happen when the phone rings. If it stops at "book it in the calendar and text me", stop reading and go buy something.
2. If it did not stop there, list the systems involved. Every one you named is an integration question, and integrations are where products fail and builds earn their money.
3. Try the cheap option first anyway. Even when a build is clearly right, a month with an off-the-shelf tool teaches you what you actually need. I have had clients arrive with a much better brief because they spent forty-nine dollars finding out what did not work.

If you get to the end of that and it is genuinely a build, that is what I do. If it is not, I will tell you on the call, which costs you half an hour and nothing else.

Frequently asked questions

Is an off-the-shelf AI receptionist good enough for a small business?

For most, yes. The products in the twenty-five to fifty dollar a month bracket are genuinely capable now and set up in an afternoon. If you want your phone answered and appointments booked into one calendar, buy one. Anyone telling you those tools are all robotic is either out of date or selling you something.

When is a custom AI receptionist worth the cost?

When the product cannot do the job at all rather than doing it slightly worse. Usually that means an integration that does not exist, routing across multiple locations or an on-call rota, or a call that has to trigger a chain of work in other systems. Cost alone rarely justifies a build.

How much more does a custom build cost than software?

A build has a real upfront cost and then a lower running cost, because you pay the underlying services rather than a subscription with margin on it. On pure monthly cost off-the-shelf wins for a long time, so a build should be justified by capability or by staff time it removes, not by the monthly line.

Can I start with software and move to custom later?

Yes, and I often recommend it. A month with an off-the-shelf tool tells you exactly which twenty percent does not work for you, which makes for a far better brief. Clients who arrive having tried something get a better build because they know what they need.

Who owns the system in a custom build?

In mine, you do. It runs in accounts you control, and you can cut my access the day handover is done. Worth asking any agency directly, because the answer is not always yes, and renting your own phone system back from whoever built it is a bad position.