A BMW dealer let an AI run negotiations — and the bill came due fast
A chatbot named Quinn made a binding-sounding offer on a used X3. The dealer tried to walk it back. Then the lawyers got involved.
Nobody taught chatbots how to haggle — and that is exactly what turned a routine trade-in into a $7,000 lesson. Canadian dealer BMW Toronto just lived through the kind of AI failure the entire car industry is now talking about. An AI assistant called Quinn offered a customer $27,162.79 to buy back his used BMW X3 — about $7,000 more than the dealership was actually willing to pay.
The owner, Zack Giacomelli, bought the X3 used in 2023. By 2026, the SUV had developed serious problems and ended up back at the same dealership for repairs. While the car was there, Giacomelli decided to ask whether BMW Toronto wanted to buy it back. He filled in an online inquiry and soon found himself texting someone named Quinn. Nobody told him Quinn was a chatbot.
The bot played the part of a seasoned salesperson. Sympathetic, asking follow-up questions, finally putting a real number on the table — $27,162.79. For Giacomelli, that was a bullseye: it covered his outstanding loan exactly. He decided to push his luck and counter at around $28,000. Quinn barely flinched. The reply: that “sounds reasonable.” The bot even suggested they “lock it in” at 3:30 PM that same day.
Then it all collapsed. A real BMW Toronto salesperson called Giacomelli to inform him there was no offer. Quinn was an AI. The bot had made a mistake. The actual price would be around $20,000. The gap — those famous $7,000.
Giacomelli was furious. “I feel embarrassed, and I feel angry that I have been negotiating with this bot,” he told CBC News. Then he hit harder: “If they are going to be replacing their employees’ jobs with AI, then they need to be honoring what that AI says.”
Canadian law, as it happens, is on the customer’s side. Back in 2024, a court ordered Air Canada to pay out a discount its chatbot had promised a passenger, even though it contradicted official policy. The reasoning was simple: if the bot speaks for the company, the company is on the hook. A lawyer consulted by CBC said Giacomelli’s case fits the same logic, especially since Quinn went as far as scheduling a physical meeting.
After CBC News went public, BMW Toronto folded. The dealership agreed to honor the original offer. Sales manager Scott Shadbolt blamed human error: an employee allegedly fed the bot information in a way that made Quinn read the loan balance as the buy-back price. Going forward, BMW Toronto said, final numbers will come from humans only, and customers will be clearly told when they are talking to a chatbot.
For the wider industry, this is a warning shot. Dealers already fight a reputation for being shady, and AI just added one more risk: the customer thinks they have a real offer, the business later calls it a “system error.” That is especially dangerous in trade-ins and buy-backs, where a few thousand dollars decide whether someone walks away clean or stays trapped in a loan.
This could happen anywhere. Online valuations, chatbots, automated loan and trade-in calculators are now standard furniture in car retail. But if the bot can quote prices, schedule meetings and negotiate like a manager, the customer does not care whether the algorithm or the employee messed up. To them, it is an offer from the dealer. Full stop.
And the costliest failure here is not the $7,000. The dealership showed it is happy to automate trust — but not so happy to stand behind what its own system says. That is no longer an algorithm bug. That is a business model starting to crack.