What to say when a prospect tried a chatbot and hated it
A prospect who says their last chatbot was awful is rejecting one specific failure, not AI. Here is how to find out which one, what to ask first, and when to walk away.
Lumina7 min read
When a prospect says "we tried a chatbot once and it was awful," agree with them, ask what it actually did, and then show them a live call on something they can test in four minutes. Do not defend AI in general. They are not rejecting AI. They are rejecting one specific thing that wasted their money, and your job is to find out which thing it was.
Most of this post is about that middle step, because it decides the deal.
What is the prospect really telling you?
Usually one of four things, and they sound identical on the phone.
The chatbot was a pop-up on the website that answered nobody's real question. Or it was an old phone tree with a new name. Or a nephew set up a tool on a free trial, never trained it on the shop's prices, and it quoted a number the owner had to honor. Or it worked fine and the owner stopped checking it.
Four different failures, four different answers. If you skip the diagnosis and go straight to your pitch, you are answering a question nobody asked.
Gartner's survey of 5,728 customers, run in December 2023, found that 64% would prefer companies not use AI in customer service, and 53% would consider switching to a competitor over it. The biggest worry was that AI makes it harder to reach a person. Your prospect has read the same headlines you have. They may have lived the number.
Why do chatbots fail small businesses?
A 2026 Berkeley California Management Review piece on chatbot frustration lists the causes. Bots misread intent. They cannot fix anything complicated. Customers get stuck in a loop with no route to a human. The piece also says bots that pretend to be human annoy people, and that generic replies make customers feel undervalued.
Read that list again as a salesperson. Every item is a thing you can test in a demo.
Does it get the question right the first time? Does it say what it is? Does it know when to stop and hand the caller to somebody? You do not need to argue any of this. You need to let the prospect poke at it.
What do I ask before I say anything about my product?
Four questions, in this order, and then stop talking.
What was it supposed to do? (Answer calls, book jobs, qualify leads?)
Where did it live, on the site or on the phone?
What was the worst thing it did?
Who stopped using it, and why?
Question three is the money. A plumber will tell you the bot gave a customer a flat-rate price for a job that needed a camera inspection. A dentist will tell you it told a patient they could be seen Saturday, and the office is closed Saturdays. Those are specific, fixable, and almost always about missing information rather than missing intelligence.
Write the answer down and repeat it back in their words. Then you have something no competitor in their inbox has done, which is listen.
How do I respond without sounding defensive?
Concede first. "That sounds like a bad setup, and I would have turned it off too." Say it flat. Prospects who expected a fight tend to lean in when they do not get one.
Then separate the failure from the category. If the old bot lived on a website, say that a website widget and a phone line are different animals. Callers in a hurry do not type. Quo analyzed over 18 million missed calls from July to September 2026 and found only 15% of callers left a voicemail, with home services at 16%. The other 85% hung up and rang somebody else. I would still check that number against the prospect's own call log before leaning on it, since Quo sells phone software and its customer base may not look like a tire shop in Dayton.
If the old bot was a phone system, ask what happened to the calls it could not handle. That answer is usually "nothing, they just dropped." Fix that story and you have a real pitch.
Should I show the demo or tell them about it?
Show it. This is the one place I will not bend. A claim that your system is different is the same sentence the last vendor used.
On the call, have them open your demo link and talk to the receptionist for the trade closest to theirs. Ask them to try the thing the old bot broke on. If it was pricing, have them ask about pricing. If it was hours, have them ask about hours on a holiday. Our platform's live demo is built for this, and a single call is capped at 5 minutes, which is plenty.
Be upfront that the demo runs against a sandbox. Bookings are simulated, and the page says so. The sandbox disclosure post covers why that helps you. Hiding it is how you get burned the day they ask to see the calendar entry.
For the full structure of that call, see the demo call playbook. If their real objection is whether callers can tell it is a machine, this post takes that head on.
When does this approach not work?
Three situations, and I would walk away from two of them.
The owner has a regulatory reason to keep every call human. Some medical and legal intake works this way. A demo will not change a compliance policy.
The owner got burned for real money and wants a refund from somebody, anybody. You are not that person, and no demo fixes anger that is really about an invoice.
The third is the quiet killer. The owner liked the old bot, it just didn't pay for itself, and they never measured anything. A better demo does not help there. You need numbers from their own phone log before you start, or you will lose to "we'll think about it."
For that last group, a free pilot with a counting rule agreed in writing gives them something to measure.
What do I price against a skeptic?
Do not discount to win a skeptic. It signals that you also doubt the product. If the platform cost is the issue, the pricing section is what you are working from: $49 a month for the platform, with client builds quoted at wholesale on a 30-minute call. What you charge the client is your decision.
What you can do is tie your first month to a number they pick. Calls answered after 5pm, say, or bookings from the weekend. Pick the number with them, not for them.
Frequently asked questions
Should I bring up the failed chatbot before the prospect does?
No, but ask early whether they have used anything like this before. If they say yes, you get the story. If they say no, you skip the whole section and save ten minutes.
What if they refuse to try the demo?
Take that as the real objection. A skeptic who won't spend two minutes testing it is telling you the problem is trust in you, or that the decision belongs to someone else. Ask who else should hear this, and offer to run it again with them on the line.
Is it honest to say my system is different from their old chatbot?
Only if you know how. Ask what the old one was built on and how it was fed information. If the difference is that your pack was written with their trade's services, price bands and objections, say that. If you can't name one, say less and let the demo do the work.
How many skeptical prospects actually convert?
I don't have a number I trust, and neither does anyone who sells this. Lumina Sales publishes no conversion data and I won't invent any. Track your own: how many skeptics took the demo, how many asked for a pilot, how many signed.
Does this apply to website chat as well as voice?
The diagnosis questions apply to both. The phone argument is easier to make because Quo has a number for callers who hang up on voicemail. I have not found a comparable source for visitors who ignore a website widget, so don't quote one.
Your next step
Before your next call, write down the three worst things you have heard a bot do. Then pick the trade closest to your top prospect, open the live demo, and try to make the receptionist do each one. Whatever it fails at is what you tell the prospect before they find it. When you are ready to put this in front of a real skeptic under your own brand, start here.
