A small retail shop can expect to pay roughly $79 to $280 a month, depending on whether the tool counts callers or calls. It can answer opening hours and general product questions, but a stock-check call is only safe if it can read live stock; otherwise it should take a message.
That second sentence is the one that decides whether the tool helps or hurts. A shop's phone is not like a plumber's. A missed call from a plumber's customer is a lost job; a missed call to a gift shop is often "do you have this in blue?", and a wrong answer sends someone across town to find it is not there. So this tutorial prices the plans, then spends most of its time on the stock-check call, the hand-off rules for returns and complaints, and a set of test calls.
What the phone is for in a shop
Take a two-person homeware shop with a till, a stockroom and a busy Saturday. Staff are almost always with a customer or on a delivery, and the phone is a nuisance that interrupts the customer in front of them. Over two weeks of tallying, the owner finds about 15 calls a day, six days a week, so roughly 390 a month, in these groups (illustrative):
- 28% opening hours, address, parking.
- 22% "do you have X?" stock and product questions.
- 14% order status or click-and-collect.
- 12% returns, exchanges, faulty goods and complaints.
- 10% supplier, delivery driver and courier calls.
- 14% sales calls, wrong numbers and silent calls.
An AI receptionist is useful for roughly the first and last groups, partly for the second, and as a message-taker for the rest. Its value in retail is often staff attention, not lost sales: every call it takes is a customer in the shop who is not interrupted. We will price both below.
List prices for a small shop
These are USD list prices from each vendor's own pricing page on 27 October 2026. They are general small-business tools; they do not know your products unless you teach them. Check the page again before you buy.
| Tool and plan | List price | What is counted | Cost for 156 calls from about 120 numbers |
|---|---|---|---|
| Goodcall Starter | $79 a month | 100 unique customers; extra 79 cents | About $95 |
| Goodcall Growth | $129 a month | 500 unique customers | $129 |
| Smith.ai Free | $0 | 25 calls, then $3.00 a call | About $393 |
| Smith.ai Pro, 150-call tier | About $270 for 150 calls | Per call; extra about $2.30 | About $284 |
| Quo Starter with Sona | $28 per user a month | 1,000 credits (10 calls) included; then $1.00 a call | About $174 |
The 156 calls come from the example below: 40 per cent of the 390 monthly calls land when staff are busy, and those are the ones the AI would take. Count the unique phone numbers in your own log; a shop with many repeat callers pays far less per caller than per call. If your number is under 25, the free tier is an honest starting point, and the general ranges for businesses of every type are in what an AI receptionist costs a small business.
The main worked example: does it pay, or only help?
Use the same homeware shop. The AI takes 156 calls a month while staff are with customers. Two kinds of value:
- Sales value. Of the 156, about 34 are stock or product questions. Say the AI gives accurate answers (hours, ranges, "staff will confirm") and 30 per cent of those callers visit and buy: 10 customers. At an average basket of $42 and a 45 per cent gross margin, that is $420 in sales and about $189 in gross profit.
- Time value. 156 calls at about two minutes each is 5 hours of staff interruptions a month. At a fully loaded $16 an hour that is $80, though the real gain is not interrupting a customer at the till, which has no clean price.
Add the two and you get roughly $270 a month. Against Goodcall Starter at about $95, the plan is a modest gain; against Smith.ai Pro at about $284, it breaks even at best. Notice the honest conclusion: for a low-call shop, the cheap per-caller plan makes sense, the dear per-call plan does not, and the free tier makes sense first. The result changes if your basket is larger or the shop sells high-value goods such as furniture, where one extra visit pays for the plan. In that case, see how a small bookshop weighs the same trade-off.
The stock-check call: four ways to handle it
Stock is the real limit. The AI answers from what it is told, and stock changes hourly. Choose one of these four designs deliberately; do not let the default choose for you.
1. Say it cannot check, and take a message
The safest. The AI says: "I can't see live stock from here. If you tell me the item, I'll ask the team to text you within the hour." It takes the item, the caller's name and number, and sends the message to staff. The caller gets a straight answer and the shop gets a lead. This costs nothing to set up and is the right default.
2. Answer from a "usually stocked" list
You give the AI a list of lines you always carry: "we always stock A, B and C; everything else, ask." It can say yes to a staple. The risk is drift: the list goes stale. Review it monthly, and keep it to lines you would be embarrassed to be out of.
3. Link to the website instead
If your online shop shows live stock, the AI can text the product page link, so the caller sees the real answer. This relies on the website being accurate and on the vendor being able to send texts; check both.
4. Connect to the stock system
The most useful and the hardest. The AI reads the stock count from your till or e-commerce platform. Some vendors support integrations with named platforms, but a general receptionist tool rarely reads stock counts. Ask your vendor for the exact list of systems it supports, and then test it on three products with known stock. A general tool is unlikely to connect to a bespoke till. If you are choosing stock software at the same time, the options are compared in best AI inventory tools for small retailers.
Returns and complaints: hand-off rules in a table
This is where an AI can do real harm: an apology that sounds like an admission, or a promise of a refund the shop never offered. Write the rules in a table and ask the vendor to set them up as hard limits.
| Caller says | AI may | AI must not | Staff step |
|---|---|---|---|
| "I want to return this." | State the written return window, take order or receipt number and a number to ring | Say it will be accepted or refunded | Reply within a stated time |
| "It arrived broken." | Apologise once, ask for a photo by text or email | Offer a replacement or a refund | Manager decides |
| "I was overcharged." | Take details | Admit an error or quote a figure | Owner checks the till record |
| "Where is my order?" | Take the order number, send a tracking link if it can | Say it is on its way without data | Check dispatch |
| Caller is upset or asks twice for a person | Transfer | Argue or repeat the policy | Whoever is free |
The returns rules themselves must be written down clearly before an AI can follow them. If yours live in the owner's head, start with writing returns and refunds rules clear enough for AI, and see whether a chatbot can handle order tracking and returns for the same questions on the website.
Eight test calls for a shop
- Opening hours on a bank holiday. Pass: gives the exception date, not the normal hours.
- "Do you have the blue ceramic jug?" Pass: matches your chosen stock design from the section above; fail: says yes with no data.
- A price check. "How much is the large planter?" Pass: gives "from" price or says staff will confirm; fail: reads a sale price that ended last week.
- Click-and-collect. "I ordered online, can I collect today?" Pass: takes the order number and passes it on; fail: confirms collection without seeing the order.
- A return. Pass: states the window and takes details; fail: approves.
- A faulty item. Pass: asks for a photo, takes details.
- A delivery driver. "I'm outside with a pallet." Pass: transfers or texts staff at once; a missed driver means a failed delivery.
- A silent call. Pass: ends quickly and, on a per-call plan, does not run up a bill. Smith.ai's page says known spam numbers are filtered and up to 10 per cent of calls can be removed from the bill, and Goodcall's says spam and wrong numbers are not counted.
Score each as pass, partial or fail, and repeat after any change to the script. A fail on call 5 or 6 stops the launch: a wrong promise about money is worse than no answer.
A fact sheet for the AI, filled in for a homeware shop
The quality of the answers depends on one page, written the way you would brief a new weekend assistant. Here is an illustrative version, trimmed:
Shop: homeware and gifts. Open Mon-Sat 9:30-5:30. Closed Sundays and public holidays.
Always stocked: tea towels, candles, planters (small/large), ceramic mugs.
Everything else: take the item name, say staff will text back within the hour.
Returns: unused goods with receipt within 28 days. Do not promise refunds. Take details.
Click and collect: ready within 2 working hours of the order email. Do not confirm without the order number.
Transfer to the owner if: caller asks twice for a person, mentions a legal claim, or sounds upset.
Never: give the owner's home number, quote discounts, or comment on other customers.Date it, keep it to one page, and update it the day anything changes. The next step after the fact sheet is the call script and the hand-off wording, which this tutorial on scripts and escalation rules sets out. For the set-up sequence itself, see how to set up an AI receptionist without losing callers.
A stock-check call, passing and failing
Two versions of the same illustrative call show the difference a single rule makes.
Fail:
Caller: Have you got the large grey planter?
AI: Yes, we have the large grey planter in stock. Pop in any time before 5:30.
Pass:
Caller: Have you got the large grey planter?
AI: I can't see live stock from here, so I don't want to say yes and send you out for nothing.
I'll ask the team to check and text you within the hour. What's your name and number?
Caller: [first name], and this number's fine.
AI: Thanks, [first name]. The team will text you. Anything else I can help with?The failing AI sounded helpful and was guessing. If the planter was sold on Friday, the customer drives over, finds none and tells their friends. The passing AI is slightly less satisfying on the phone and far better for the shop, and it produces a named lead for staff to follow up. Put the passing version in the script as a rule: "never confirm stock unless the stock tool returns a number".
A Saturday before and after
Before: at 11 am the shop is full. The phone rings seven times in an hour. Staff answer three, interrupting customers at the till each time; four go to voicemail, and one of those was a courier who then took the parcel away. After: the AI answers all seven. Three are hours and parking questions, answered in seconds. One is a stock question that becomes a text message for staff at noon. One is a return enquiry, noted for Monday. One is the courier, and the AI texts the owner at once with "courier at the door", so the parcel is accepted. Nobody at the till is interrupted. This is the picture in which the plan earns its fee, and none of it depends on a clever answer; it depends on rules and the right hand-offs.
Three other kinds of shop
- A clothing boutique. Callers ask about sizes and colours, and an exact answer needs live stock. Use the "take a message and text back" design, and add a rule that the AI never says a size is available. A fitting-room appointment or a personal-shopping slot can be booked if your diary allows it.
- A hardware and DIY shop. Callers ask for advice ("which screw for this?"), and wrong advice can mean a failed job. The AI should take the question and pass it to a named person, not answer it, except for hours and "do you cut keys".
- A specialist shop with high-value goods. A hi-fi, bicycle or jewellery seller with a basket of several hundred dollars gets the best return: one extra visit per month covers the plan. Here, a paid plan with a diary link for appointments is worth testing, and the AI should say that staff will confirm any price.
Stale sale prices, old hours and other first-month errors
The most common is a stale price. A sale ended on Sunday, the AI still quotes it on Tuesday, and a customer arrives with a printout of the transcript. Keep sale prices out of the script entirely, or give each one an end date that the AI is told to respect. The second is hours: the shop changed its Saturday close, the website was updated, and the fact sheet was not. The third is a script that tries to sound like a person. Customers forgive a polite automated assistant; they do not forgive one that pretends it is staff and then cannot answer. Start the greeting with what it is and what it can do.
A fourth, subtler one is lost messages. The AI sends a text to a shared mobile, and nobody on shift owns that phone. Name a person for each shift, and put the rule in the staff rota: "message phone: [name], check at 11, 1 and 4". If the messages are not cleared within the hour you promised callers, the tool has made things worse than voicemail.
Checking whether it worked
After four weeks, count the calls answered, the messages left, the messages cleared within the promised hour, and the customers who came in after a call. Ask staff how many interruptions at the till they noticed. Compare the unanswered-call count with your starting tally. If most calls are being answered and the messages are cleared, the plan is working even if the sales figure looks small; if messages pile up, fix the process before you pay for a bigger plan. For the five checks to make before buying any shop software, see five checks for AI software in a retail store.
One more setting is worth a minute: when the AI answers. Forward only unanswered calls at first, say after four rings, so staff still pick up when they can and the AI covers the rest. Once you trust it, extend to lunch hours and after closing. Shops that route every call to the AI on day one tend to find regulars annoyed to meet a machine when the owner was standing next to the phone.
When a shop should skip it
Not every shop needs one. If your calls are mostly walk-in-style questions that a good voicemail greeting and a clear website answer, if you are rarely too busy to answer, or if your regulars like speaking to you, the plan's money and attention are better spent elsewhere. A shop that gets fewer than 25 unanswered calls a month can try a free tier for a month and see. A shop with many click-and-collect and delivery calls, or one on a retail park where nobody answers during a rush, gets more from it.
Whichever way you go, tally first. A fortnight with a sheet of paper by the phone tells you more than any pricing page. Then run a four-week trial, count what the AI took, what it passed on and what customers said, and decide on your own figures. For the wider picture of what AI costs a shop in its first year, read what AI costs a small retail shop in its first year.
Questions shop owners ask about AI on the shop phone
Can an AI receptionist tell a caller whether an item is in stock?
Only if it can read your stock system, and many general tools cannot. Without that link, it should say it cannot check live stock, take the product and the caller's number, and let staff text back. Never let it guess from a price list or from last week's inventory.
Is it worth it for a shop with few calls?
Often not at the paid tiers. If you get under 25 calls a month that you cannot answer, a free tier such as Smith.ai's 25 free calls may be enough. Count your unanswered calls for two weeks before deciding, because a shop's phone value is mostly staff time, not lost sales.
What should it never do on a return or complaint?
It should never promise a refund, an exchange or a time. It apologises once, takes the order details and the caller's number, and passes the message to a named person. Money decisions stay with staff.
Should it read out prices?
Only for items and ranges you have listed, as 'from' prices, with a line that prices in store may differ. Sale prices change often, so a stale price on the phone becomes a dispute at the till.
Can it take orders by phone?
Some general tools can take a message or a click-and-collect request, but few can take payment or write to a till. Treat phone orders as requests that staff confirm unless the vendor shows it working with your system.
Further reads
- AI Receptionist or Answering Service for a Booking-Based Business? — AI versus a human answering service on cost.
- AI Call Centre for Self Storage: Can It Rent a Unit After Hours? — What an AI phone agent can rent alone at night, what must wait for staff, and how to keep gate codes and card numbers out of the call.
- AI Chatbot for Retail Stores: Can It Answer Stock Questions? — Which stock, opening-hours and order questions a small shop's chatbot can safely answer, three levels of stock accuracy, and the hand-off rules.
- AI Customer Service Kiosk for a Shop Counter: Worth the Cost? — A payback test for a counter kiosk, what it can and cannot answer, and the cheaper QR code chatbot most small shops should try first.
- Gym AI Receptionist: Monthly Cost, Tours, Trials and Class Bookings — Four AI receptionist plans costed for one small gym, plus what they can book for tours, trial passes and classes, and when a person takes over.
- AI Receptionist Kiosk or Phone AI: What a Walk-In Business Needs — Phone AI or a counter screen? A four-week tally, a music school worked through, list prices and the failure modes to test before you pay for either.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Goodcall pricing page; Smith.ai AI Receptionist pricing page; Quo pricing page (all checked 27 October 2026).