Per minute
Standard for voice AI, because a call has a duration. Cheap when calls are short and brutal when a caller rambles. Ask how partial minutes are rounded.
READ THE QUOTE PROPERLY
AI receptionist pricing comes in four shapes: per minute, per conversation, per reply and per seat. Two honest vendors can quote wildly different numbers purely because of the unit, so the useful question is not what does it cost but what counts as one of those.
Standard for voice AI, because a call has a duration. Cheap when calls are short and brutal when a caller rambles. Ask how partial minutes are rounded.
Common in messaging. Ask exactly when a conversation starts, when it ends, and whether a customer returning the next day starts a new one.
The most predictable unit if you know your inbound volume. Ask whether confirmations, menu taps and handover notices count.
Priced by staff logins rather than customer contact. Fine for a team inbox, expensive if the owner needs a seat purely to watch.
| Driver | Why it moves the number | What to do about it |
|---|---|---|
| Inbound volume | Every unit above is ultimately a count of customer contact. | Pull last month's real enquiry count and quote against that, not a round number. |
| Conversation length | Per-minute and per-conversation pricing both reward finishing quickly. | Ask what the vendor's average is for a business like yours, and whether they will show you. |
| Outbound messages | Reminders and follow-ups start conversations, which is where charges usually apply. | Decide whether you want reminders at all before including them in the comparison. |
| Number of staff | Seat pricing multiplies with the team, not with the customers. | Count who genuinely needs a login rather than who might like one. |
| Integrations | Calendars, practice systems and CRMs are where custom work hides. | Name every system you expect it to touch, in the first conversation. |
| Seasonality | A quiet January and a brutal December average out badly on an annual plan. | Ask what happens in your peak month, not your typical one. |
Three of them, and they are the reason quotes stop matching invoices.
First, messaging fees. On WhatsApp, Meta charges by message category and country. Replies inside the 24-hour customer service window are not charged under Meta's current published rules, but reminders and re-engagement messages are. Ask whether the vendor bills these through or whether you attach your own payment method.
Second, telephony. Voice products need phone numbers and minutes, and those are usually separate from the AI itself.
Third, setup. Somebody writes down your services and prices, checks what the system learned and connects the calendar. If a vendor says setup is free, ask who is doing it, because the honest answer is usually you.
Ask all nine, in writing, of every vendor on your list.
This comparison is usually framed as a salary against a subscription, which flatters the software and misses the point. A person can read a room, calm someone down, spot that a regular is unhappy, and do six other jobs between calls. Software cannot do any of that.
What software does is cover the hours nobody is on duty, and absorb the repetitive half of the enquiries without getting tired or short. In most small businesses the honest answer is not one or the other, it is a person who is no longer interrupted forty times a day.
Searching for reviews of this category is unusually unhelpful, and it is worth knowing why before you spend an evening on it. Threads on Reddit and the trade forums rarely say which channel the product used, what volume the business had, which year it was, or which tier they were on. A dental practice's experience of a voice product in one country tells you very little about a salon's experience of a messaging product in another.
The substitute is cheaper than it sounds: run your own two-week test with real customer messages. Take last month's genuine enquiries, send them through, and count how many were finished rather than acknowledged. That is a review of your business, which is the only one that predicts anything.
Elily is priced by AI replies rather than by minutes or seats, because the unit people can actually count is how many customer messages get answered.
Free includes 150 AI replies a month, with no expiry and no cap on the bookings those replies produce. Lite and Pro raise the allowance, remove Elily branding from Lite upwards, and add reminders, follow-ups, approval mode and multi-staff calendars on Pro.
Prices are shown in the currency of the country you are browsing from, and Meta's message fees, where they apply, are separate and paid at the rates Meta publishes for your country.
Start on Free, send Elily your website, and let a real month of customer messages produce the number. It beats a spreadsheet built from list prices.
QUESTIONS BEFORE YOU START
It depends on the pricing unit far more than on the vendor. Voice products usually charge per minute, messaging products per conversation or per reply, and team products per seat. Quote every vendor against your real inbound volume from last month and compare the totals, not the headline rates.
Per hour, almost always. That is the wrong comparison for most small businesses, because a receptionist does many other jobs. The realistic framing is what it costs to cover the hours nobody is on duty, and to stop the interruptions during the ones you are.
Usually three: messaging or telephony fees charged by the platform rather than the vendor, setup work that somebody has to do, and integrations quoted as projects. None of them are dishonest, but they are often absent from the first number you are given.
There are free plans, as opposed to trials. Elily's Free plan is 150 AI replies a month with no expiry and no booking cap. Check whether a free offer is a plan you can stay on or a countdown to a card form.
KEEP READING
THE NEXT MESSAGE CAN BE HERS
Start on Free and let a real month of enquiries tell you what this is worth before you compare a single price list.
Last updated