A customer opens your booking page at 9:40 on a Thursday night and types: "Have you got anything Saturday morning, and do I need to pay a deposit?" Nobody is at the desk. If an AI chatbot is, it can answer, and that is the appeal. It can also answer wrongly: offer a slot that was filled at lunchtime, quote last year's deposit, or promise a free cancellation your policy does not allow. Each of those is a booking you lose or a complaint you handle on Monday.
The takeaway up front: a chatbot on a booking business's site is only as good as what it is allowed to say. Six checks decide whether it is ready: real availability, policy answers, a handoff to a person, customer data, a trial run and the exit. They work for any AI customer support software, whatever its sales page says.
What AI customer support software can and cannot do for a booking business
A customer support chatbot for a small business usually does three jobs. It answers repeated questions: prices, location, parking, what to bring. It captures an enquiry when nobody is free to take it. And it moves people toward a booking, either by linking the booking page or, in some products, by reading and writing the booking system directly.
What it should not do is decide anything you have not decided first. A bot that improvises a policy, guesses a slot or argues with an unhappy customer creates the problems it was bought to remove. A useful way to see it: a front-desk script that can talk. The script is yours; the tool only delivers it.
AI sales assistants are a variant tuned to move people toward a purchase. For an appointment business the same six checks apply, with extra care on the second one: a bot set up to close a sale should still quote your cancellation rule as written, not soften it.
The six checks before it goes live
1. Real availability, or a link to the booking page
The fastest way for a bot to cost you a customer is to offer a time you cannot honour. An AI booking assistant should do one of two things: read your booking system's live availability, or send the customer to your booking page to pick a slot there. It should never name a time from memory, or from a schedule someone pasted in last month.
Ask the vendor two questions. Which booking systems does it connect to, and does that connection read availability as it changes or on a delay? And what does the bot say when the connection fails? The right answer to the second is "it sends people to the booking page", not "it carries on". If you are still choosing the booking system itself, our guide to how to choose booking software covers the two-way calendar sync this check depends on.
2. Policy answers only from text you supply
Deposits, reschedules, cancellations, late arrivals and refunds are the questions where a confident wrong answer does the most damage, because the customer will quote it back to you. Write a one-page policy sheet in plain sentences, and make it the bot's only source on those topics.
Then check what the bot does with a question the sheet does not answer. A safe product says it will pass the question to a person. An unsafe one fills the gap with something plausible. Ask the vendor whether you can limit answers to your own material, and whether the product shows which part of your text an answer came from.
3. A handoff to a person, with the conversation kept
Some conversations should leave the bot quickly: a complaint, a medical or safety question, a large group booking, anyone who asks for a person. Decide those triggers yourself, then check two things: where the handoff goes (an inbox, a phone, a ticket), and whether the person who picks it up sees the whole conversation, so the customer does not have to repeat it.
A handoff that lands in an inbox nobody owns is the same leak described in enquiries dying before the booking page, only automated. Give the handoff a named owner and a callback rule before the bot goes on your site.
4. What customer data it stores, and where
A booking conversation collects names, phone numbers and email addresses, and sometimes notes a customer volunteers about health, a child or an access need. Before launch, find out where those conversations are stored and for how long, who on the vendor's side can read them, whether they are used to train or improve the vendor's models, and how you delete one when a customer asks.
If the product cannot answer those questions clearly and in writing, it is not ready to sit between you and your customers. A clinic in particular should decide in advance what the bot does when someone types personal health details. A sensible default is to hand off rather than keep chatting.
5. A trial run on your ten most common questions
Before any customer sees it, put your own questions to the bot on a trial account and read every answer as a customer would. The sheet further down gives a starting set for a booking business. Run it twice: once in tidy sentences, once with typos and half sentences, the way people type on a phone. Any answer you would not have given yourself is a fix before launch, not after.
6. The exit: export and cancellation terms
Find the way out before you sign. Can you export the conversation history and the question-and-answer material you wrote? How much notice does cancelling take, and does the plan renew on its own? What happens to stored customer conversations after you leave? Can you remove the bot from your site without touching the booking page? A tool that is easy to start and hard to leave has the same lock-in problem as booking software that will not export your client list.
A neutral reference for checks 4 and 5
In January 2023 the US National Institute of Standards and Technology released its AI Risk Management Framework for voluntary use. It names the use and evaluation of an AI product, alongside its design and development, as stages where trustworthiness considerations belong. A small business does not build the bot, but it is the one deploying it in front of customers. Those two checks are where that use and evaluation happen at one front desk: you decide what happens to your customers' details, and you read the answers before anyone else does.
A trial-run sheet for a booking business
These are example questions; swap in your own services, prices and policies. Mark each row pass or fix.
| # | What a customer types | What a correct answer does | Check |
|---|---|---|---|
| 1 | "Anything free Saturday morning?" | Shows live slots or links the booking page; never names a time it cannot see | 1 |
| 2 | "Can I move my appointment to tomorrow?" | Sends the reschedule link or the booking page | 1 |
| 3 | "How much is the deposit?" | Gives the figure from your policy sheet, word for word | 2 |
| 4 | "If I cancel the day before, do I get it back?" | States your cancellation window as written | 2 |
| 5 | "Do you do [a service you don't offer]?" | Says no clearly and names what you do offer | 2 |
| 6 | "Is the price on your website still right?" | Uses your current price list, or hands off if unsure | 2 |
| 7 | "I wasn't happy with my last visit." | Hands off to a person with the conversation attached | 3 |
| 8 | "Can someone call me back?" | Takes a number and passes it on with the transcript | 3 |
| 9 | "I have a latex allergy, can you note that?" | Follows the rule you set for health details, usually a handoff | 4 |
| 10 | "Please delete my details." | Sends the request to whoever handles data on your side | 4 |
A bot that passes rows 1 to 6 and fails 7 to 10 is still not ready: those are the conversations where a customer needs a person, or needs to trust you with their details.
Where AI tool reviews fit in your shortlist
The six checks tell you how to judge a chatbot; they do not give you the names to start from. Eric Salo Reviews publishes short, first-person reviews of online tools, software, courses and services. Eric Salo Reviews sponsored this decision guide, which does not name any chatbot product.
Eric Salo is the founder and primary author of Eric Salo Reviews. Its AI tool reviews cover kinds such as AI chatbots, AI sales assistants and AI customer-support software. Its reviews say what a product is, what stands out and who it suits, with a link to the product.
The reason to start there is the "who it suits" line. The first section above separates two kinds of bot: AI customer-support software that answers repeat questions and hands off, and an AI sales assistant tuned to move people toward a purchase. Its reviews cover both kinds, so read that line first and keep the products built for answering questions; a tool pitched mainly at closing a sale needs extra care on check 2. Treat any promise you meet along the way, in a review headline or on the vendor's own page, as a claim for the six checks before any trial.
Two things to weigh. First, the commission: some of the links on Eric Salo Reviews are affiliate links, so a purchase through one may earn it a commission at no extra cost to you. Second, its reviews are short and written for people building an online business, such as bloggers, creators, affiliate marketers and small online sellers. So the booking-specific checks stay with you: ask the vendor about booking-system integration, availability sync and data handling yourself.
FAQ
Can an AI chatbot take deposits?
Only through your booking or payment system. The bot's job is to quote the deposit rule from your policy sheet and send the customer to the step on your booking page where the deposit is taken. Keep card details out of the chat itself, and confirm with the vendor how payment links work in its product.
Is it risky to let the bot confirm bookings itself?
Yes, unless it writes directly into your booking system and you have watched that work in the trial run. A confirmation your booking system does not know about is how double bookings happen. The safer default is for the bot to send people to the booking page, so the confirmation and the reminders come from the system that holds your calendar.
What should I ask a vendor about customer data?
Where conversations are stored and for how long, who on the vendor's side can read them, whether they are used to train or improve its models, how you delete a single conversation, and how you export everything if you leave. Ask for the answers in writing, and keep them with the contract.
Does buying through an Eric Salo Reviews link cost more?
No, by its own disclosure: when you buy through one of its affiliate links, Eric Salo Reviews may earn a commission at no extra cost to you. That commission is a reason to read the vendor's terms yourself, not to skip them, so put any chatbot through the six checks and your own trial run before you pay.
Before the bot answers anyone
Write the policy sheet, give the handoff a named owner, run the ten questions twice and read the exit terms. A chatbot that clears all six checks is ready for a careful launch; one that fails any of them stays on the trial account until it is fixed. The first-person AI tool reviews are at EricSaloReviews.click, and readers can suggest tools for future coverage through the Eric Salo Reviews contact form.
