"Receptionist software" covers at least four different products that share almost nothing. Working out which one you are shopping for saves a great deal of time, because the vendors in each category do not compete with each other despite using the same words.
The four categories
Visitor management. A tablet in a lobby where people sign in, print a badge, and notify whoever they are visiting. Office product. Nothing to do with the phone.
Front-desk practice software. Scheduling, client records, payments — the system a receptionist sits in front of all day. Dentists, clinics, salons. This is software for a receptionist rather than software that is one.
Call centre and console software. A screen showing incoming calls with buttons to route them. Software for a person whose job is answering a switchboard.
Software that answers the phone itself. An AI receptionist. No person in the loop for the calls it handles.
Those are four different purchases. A lot of confusion in this market comes from searching the category name and getting a list that mixes all four.
Which one do you need?
Work backwards from the problem.
People arrive and nobody knows they are there — visitor management.
Your front desk is drowning in admin — practice software.
A person answers a high volume of calls and routes them — console software.
Calls go unanswered because nobody is free — an AI receptionist, and this is by far the most common problem for a salon or a restaurant.
The last one is worth dwelling on. It is not an efficiency problem, which is what the other three solve. It is a coverage problem: the calls are not being handled slowly, they are not being handled at all.
If it is the phone problem
Then the question that matters is narrower than any feature list.
Can it complete a booking, or does it take a message?
Roughly half of what is sold as receptionist software cannot write into your calendar. It answers, it converses, it hands you a message. For an appointment business a message is the beginning of work rather than the end of it — someone has to read it, check the book, ring back, and hope the caller has not booked elsewhere.
Ask it first, get it in writing, and most of the confusing middle of this market resolves itself.
The features that actually matter
Your real service list. Durations, prices, who can do what — read from the system you actually book in, not typed into a form at onboarding and left to go stale.
A hold between offering and confirming. The seconds while a caller decides are where double bookings come from. Preventing double booking has the detail.
Naming a booking before changing it. When a caller has more than one appointment, the system must say which one it found before touching anything.
Knowing when to stop. Allergy questions, complaints, anything medical, anything priced by inspection. A system that improvises here is a liability, not a convenience.
Readable transcripts. So you can find out what your callers actually ask, which is reliably not what you assumed.
The features that do not
Voice cloning. Callers work out it is software within a sentence anyway, and mostly do not mind.
A long integration list. You need one integration — the calendar you actually use.
Outbound calling. Different product, real legal obligations in the United States around consent and calling hours, and not something to switch on casually.
Sentiment dashboards. For a five-person business this is a chart nobody opens.
What it costs, and the model matters more than the number
Per minute, per call, per booking, or flat. A service quoting fifty dollars can cost more than one quoting two hundred depending entirely on your volume, and usage-based pricing means your busiest month is your most expensive.
We broke the four models down, with the fees that sit outside the headline rate, in what an AI receptionist costs.
How to test before committing
Ring whatever you are considering and try three things: ask for a vague time, interrupt it mid-sentence, and ask it something it cannot know.
The third is the one that matters. A system that invents a price or reassures someone about an allergen will eventually do it to a real customer, and it will sound entirely confident while doing so. The full set of five tests, all of them failures we found in our own system before a customer met them, is in five calls to make before you trust an AI receptionist.