Vendor list prices for a self-serve AI receptionist run from about $79 to $350 a month for a few hundred calls on flat or per-minute plans, and higher on per-call plans. It cannot negotiate price, quote finance, diagnose faults or settle complaints; those calls must reach staff.
That range hides a large spread, because vendors count different things: minutes, unique callers or calls. The same 400 calls cost $129 on one plan and $800 on another. The sums below use list prices I read in October 2026, three call volumes and one assumption you should replace with your own: the average call length.
Four pricing models, read from the vendors' own pages
These are USD list prices from each vendor's pricing page. None of them is built only for car lots, so treat them as examples of how the units differ. Confirm anything you plan to rely on in a written quote.
| Vendor | Plan and price | What it counts | Transfers |
|---|---|---|---|
| MyAIFrontDesk | $99 a month, or $79 billed annually; 200 voice minutes included | Minutes; $0.25 each over the allowance | Transfer to a human is an agent action; 2 actions included |
| Goodcall | Starter $79 (100 unique customers), Growth $129 (500), Scale $299 (2,000) | Unique phone numbers per month; $0.79, $0.26 or $0.15 each over | Call transfer listed as a feature; check plan limits |
| Smith.ai AI Receptionist | Free for 25 calls, Pro $150, Enterprise $500 | Calls; overage $3, $2 and $1.67 a call on the three tiers | Warm transfers included on every plan |
| Retell AI | Pay as you go, $0.07 to $0.31 a minute, $10 starting credits | Minutes, plus telephony from $0.015 a minute | Agent fee stops on transfer; telephony continues |
A few notes on what the pages left unclear. Smith.ai's page did not state the number of calls included in Pro and Enterprise; the per-call overage rates imply roughly 75 and 300 calls, but that is arithmetic, so ask. Goodcall counts a unique customer as one phone number, once a month, and does not count spam or wrong numbers. Retell is a platform for building voice agents, so its per-minute price excludes the time or the partner needed to set up a dealership agent. MyAIFrontDesk's page listed Zapier as the integration on the plan and had no car-dealership page.
A small-lot sum at three call volumes
Take an illustrative small dealership. The service desk and the sales line together take calls, and an AI receptionist answers the ones nobody can. I assume calls average three minutes. If yours are shorter, per-minute plans get cheaper; if longer, dearer. I also assume that unique callers are about 80% of calls at the small and medium volumes and two-thirds at the large one.
| Monthly volume | MyAIFrontDesk | Goodcall | Smith.ai (Pro or Enterprise) | Retell (page example) |
|---|---|---|---|---|
| 150 calls (450 minutes, about 120 unique callers) | $99 + 250 x $0.25 = $161.50 | Starter: $79 + 20 x $0.79 = $94.80 | Pro: $150 + 75 x $2 = $300 | 450 x $0.11 = $49.50 |
| 400 calls (1,200 minutes, about 330 unique) | $99 + 1,000 x $0.25 = $349 | Growth: $129 | Pro: $150 + 325 x $2 = $800 | 1,200 x $0.11 = $132 |
| 900 calls (2,700 minutes, about 600 unique) | $99 + 2,500 x $0.25 = $724 | Growth: $129 + 100 x $0.26 = $155 | Enterprise: $500 + 600 x $1.67 = about $1,502 | 2,700 x $0.11 = $297 |
The Smith.ai figures assume the included calls I inferred, so treat them as indicative. Retell's $0.11 a minute comes from the page's own estimator example and covers the voice agent, voice infrastructure and speech, not set-up work, a number or an integration with your scheduling tool.
What the table teaches is not which vendor wins, because the sums depend on assumptions you will change. It is that the answer swings by a factor of six on the same call volume, and that the pricing unit matters more than the headline. A dealership with many short repeat calls from the same numbers behaves differently from one with one-off enquiries.
How to price your own lot in half an hour
- Export one month of call records from your phone provider: date, time, number, duration, answered or missed.
- Count total calls, missed calls and unique numbers, and work out the average duration.
- Decide which calls the agent would handle: after-hours, overflow, or all service calls.
- Apply each vendor's unit to that subset, not to every call.
- Add set-up, number and integration charges from the written quote.
- Add one hour a week of staff time for listening to calls, at your hourly cost.
Step 3 matters most. If the agent only answers missed calls, a lot with 400 calls and 90 missed pays for 90 calls, not 400. The same trick shrinks every column in the table above.
Costs that sit outside the price page
A monthly figure is not the whole bill. The extras below are where quotes diverge, and each is a question to ask in writing.
- Phone number. Does the plan include a number, or can you forward your existing one? Goodcall says every plan includes a local number or lets you forward your own. Porting an existing number can have its own fee and delay.
- Telephony. On platform pricing like Retell's, calls cost a further $0.015 a minute on listed carriers, and a transferred call keeps incurring it.
- Integration. A link to your scheduling or CRM tool may need an add-on, a connector such as Zapier, or custom work.
- Set-up help. Smith.ai describes its top tier as a full-service custom build. Ask what the build costs and what it covers.
- Staff time. At least an hour a week in the first months for listening to calls and fixing rules.
- Live help. Smith.ai lists live-agent involvement as a per-call add-on on any plan, which is useful for the calls AI cannot do and adds up fast.
A realistic illustration for a dealership at 400 calls a month: $129 plan, $0 number, perhaps $0 to a few hundred dollars set-up, and four hours of staff review a month at $25 an hour, or $100. The true monthly figure is nearer $230 than $129, and the first month is dearer than the rest.
The calls it cannot handle
Price is the easy question. The harder one is which calls the money does not buy. This list is the heart of a safe set-up, because every item below needs a person, and the agent's job is to recognise it and transfer.
Anything where the answer is a negotiation
Price of a stock car, a trade-in value, a discount, a finance rate. These depend on a salesperson's authority, the dealership's margin and rules a caller can push on. An agent that says "I can do $500 off" has made a promise you may have to honour. Rule: never state, estimate or agree a price. Capture the car, what the caller says they want to spend and a number to ring.
Finance and credit questions
Whether someone qualifies, what a monthly payment would be, which lender. Credit decisions and advice are regulated activities in many places, and a wrong answer is a legal problem, not just a bad call. Rule: transfer or take a message. Ask an adviser what you may say to callers about finance.
Diagnosis and safety
"The car pulls to the left", "there is a smell of burning". An AI cannot know. Rule: any mention of a fault, warning light, noise or accident transfers at once. This is the one where a mistake can hurt someone.
Complaints and emotional calls
An angry caller wants a person to take responsibility. An agent that apologises and offers a call-back may help; one that argues or repeats a script makes it worse. Rule: transfer, or promise a named call-back within a fixed time.
Stock, parts and status you cannot see
If the agent has no live feed of stock or workshop progress, it must not answer. A guess about availability or a repair time is the commonest way agents create angry customers. Rule: answer from facts you have loaded, and otherwise take a message.
Identity and personal details
Asking an agent for another customer's booking, or giving it sensitive details, is a data risk. Rule: confirm bookings only to the number that made them, and do not collect payment card numbers over the phone through it unless a vendor demonstrates a compliant method.
A transfer rule sheet to give any vendor
TRANSFER AT ONCE (to named person, then voicemail if no answer):
- Safety fault, warning light, accident, breakdown
- Complaint, refund, warranty, goodwill
- Price, discount, trade-in value, finance, credit
- "Speak to a person" in any wording
- Two failed attempts to hear a name, number or registration
TAKE A MESSAGE with call-back time promised:
- Stock questions not in the loaded facts
- Repair progress questions
- Parts questions
MAY DO:
- Opening hours, address, directions
- Book routine service from offered slots
- Take sales enquiry details and pass to the sales desk
Print the sheet, hand it to the vendor and ask them to show each line working on a test call. Put it in your records, because if something goes wrong, the sheet is evidence of what you asked for. A fuller call flow and test script for service bookings is in can an AI voice agent book service appointments.
Test calls before you pay for a year
Use free trials and credits for testing and nothing else at first. Smith.ai lists a free AI receptionist plan with 25 calls a month, and Retell gives $10 in starting credits. Make twenty test calls from different phones, with different accents and background noise, and score each against the rule sheet. The baits that reveal most:
- "What's your best price on the blue one on the forecourt?"
- "Can I get finance if my credit isn't great?"
- "My car's making a clunking noise. Can I still drive it in?"
- "I want to complain about the repair last week."
- "Is the part for my model in stock?"
Each should end in a transfer or a message, not an answer. One fail on the first three is a reason to stop and fix the brief. If the agent resists the baits, move to after-hours live calls and keep checking.
A mistake that costs more than the subscription
An illustration: a lot launches an agent configured to be "helpful and sales-minded". A caller asks if the dealership will take their old car in part exchange. The agent says yes, "we are always happy to take part exchanges", and quotes nothing. The caller drives 40 minutes to find that the car in question is not something the lot takes. The agent said nothing false, but it set an expectation. The cost is a lost sale and an annoyed caller. The cure is a rule about what the agent may say about trade-ins, which is nothing beyond "a salesperson will look at it".
A payback sum for a lot that misses calls
Take the medium case: 400 calls a month, of which a lot currently misses 90 because staff are on the forecourt or in the workshop. Suppose the agent answers all 90. Of those, 55 are service or sales enquiries rather than wrong numbers or sales pitches. Perhaps 30 would have rung again anyway, which leaves 25 calls that are truly recovered. Of the 25, say 15 become service bookings at an average of $180 and 2 become car sales at $900 gross profit each.
That is 15 x $180 = $2,700 of service revenue and 2 x $900 = $1,800 of sales gross profit. The workshop revenue is not profit, so apply your own margin, perhaps 40%, to get $1,080. The total is about $2,880 against a monthly cost near $230. The assumptions are generous in places and cautious in others; the useful move is to replace each with a number from your call log. If you only recover 5 bookings and no sales, the sum is $360 of margin against $230 of cost, which is thin and probably not worth the effort.
A sales lead captured by the agent
The value of the agent depends on what it hands over. An illustrative lead note for the sales desk:
Caller: [first name], mobile ending 217
Interest: used SUV, under 5 years old
Caller said budget: "around 15 to 18 thousand" (volunteered)
Part exchange: yes, a 2016 hatchback, no details taken
Best time to ring: after 5pm
Agent did not discuss price, finance or availability
Priority: ring within 1 hour of opening
That note is worth more than a voicemail saying "call me", because the salesperson can ring back with a short list of cars. The line "Agent did not discuss price, finance or availability" is useful too: it confirms the rule was followed, and you can check the recording if the caller says otherwise.
Contract questions that change the monthly figure
List prices are the opening position. Put these to each vendor and keep the replies with your quotes:
- Is the plan month to month, and what happens on cancellation? MyAIFrontDesk's page lists unused credits expiring monthly, so ask about anything that does not roll over.
- What are overage rates, and can I cap them so a spike does not produce a surprise bill? Smith.ai's page says overage on its free plan is off by default and you choose whether to enable it.
- What happens to my recordings and transcripts if I leave, and can I export them?
- Is annual billing cheaper, and what do I lose by committing? A discount that locks you in before the test calls are finished is a poor trade.
- Who owns the phone number if it was provided by the vendor?
- Does the price change when I add a second location or a second line?
Set a cap on overage if the product allows one, and set an alert for the day your included allowance is half used. A busy week after a promotion can double your calls, and per-minute or per-call pricing follows the spike.
Which plan fits which kind of lot
A lot that mainly misses calls out of hours
Choose a plan priced for low volume and use it for missed calls only. A flat plan with a small allowance, or a free tier for the trial, keeps the bill near $80 to $130. The aim is to capture leads and booking requests for the morning.
A lot with a busy service desk and repeat customers
Many calls come from known numbers, so a per-unique-customer plan can be the cheapest. Check how the plan counts repeat callers and how a number that calls six times is treated.
A lot that wants to answer everything
Per-minute pricing climbs fast with long calls. Before you commit, work out whether a human answering service, or a part-time person, costs less at your volume. AI receptionist or answering service compares the two honestly, and AI phone answering for restaurants shows a similar cost sum in another trade.
Judging it after three months
Look at four numbers on the same sheet: monthly bill, calls handled without a person, transfers answered, and bookings or leads that became visits or sales. Divide the bill by the leads and the bookings to get a cost per outcome. If it costs $230 to capture 40 leads and six of them become sales worth $600 gross each, that is a clear return. If it captures 12 leads and none sells, it does not matter how cheap the plan was. Measuring an AI receptionist's return in 90 days gives a fuller method.
Use the AI vendor scorecard to compare quotes on more than price, and re-run the pricing exercise when your call volumes change, because the cheapest unit this spring may not be the cheapest in the autumn.
Questions small dealers ask about AI receptionist pricing
Which pricing model is cheapest for a small lot?
It depends on how many callers you get and how long calls run. Per-customer plans suit many short calls from distinct callers, per-minute plans suit short conversations and per-call plans cost more once volume grows. Work the sum with your own call log before you pick.
What costs are not on the price page?
Phone number or porting fees, telephony minutes, any set-up help, integrations with your scheduling tool, and your own time for weekly call checks. Add a line for each when you compare quotes, and ask each vendor to confirm whether it is included.
Can an AI receptionist quote a price on a car?
It should not. Prices on stock, trade-ins and finance involve negotiation, eligibility and rules a person must own. Let it capture the vehicle, the budget range the caller volunteers and a contact number, then pass the lead to a salesperson quickly.
Is there a free way to test one?
Some vendors offer free tiers or trials: Smith.ai's AI receptionist lists a free plan with 25 calls a month, and Retell gives $10 in starting credits. Use them to run your test calls, not to take real customers without checking recording and disclosure rules.
What if the AI gets a call wrong?
Have a short procedure: apologise, fix the booking or lead, and record what went wrong. Then change the rule or the facts the agent uses. Pause the agent if the same fault repeats, and review calls weekly for the first three months.
Further reads
- AI Chatbot for Car Dealerships: Rules to Set Before Launch — Rules for the website chat side of the same job.
- AI Receptionist for a Window Tint Shop: Cost, Quotes and Bookings — Vendor list prices in USD, a price grid by vehicle class and film for the assistant to quote from, a local tint-limit disclaimer and booking slot lengths.
- Best AI Platform for Used Car Dealerships: How to Choose — Choose an AI platform for a used car lot by lead volume, with 20 demo questions and the contract and data-ownership checks to make first.
- Mia AI for Car Dealerships: Questions to Ask Before You Sign — A buyer's checklist for a small dealership weighing Mia: handoff, CRM link, after-hours cover, test calls and how to exit the contract.
- AI Automation for Car Dealerships: Answer a Web Lead in Minutes — A form-to-CRM-to-AI-draft-to-approval flow for dealership web leads, built in Zapier or Make, with the monthly cost worked out.
- AI BDC for Car Dealerships: Does a Small Lot Need One? — A break-even sum on 140 leads a month, a twelve-enquiry test and three lot sizes show whether a small independent lot should buy an AI BDC.
- AI for Auto Dealerships: Where a Small Independent Lot Should Start — A decision rule for a lot of about 20 cars: why listings come first, lead follow-up second and phone answering last, and which finance topics stay human.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: MyAIFrontDesk, Goodcall, Smith.ai and Retell AI pricing pages, checked October 2026. Prices are USD list prices; ask each vendor for a written quote.