Air Canada and the chatbot that promised a discount: a company answers for what its AI says

The story is simple, which is exactly why it works so well as a lesson. In November 2022, following the death of his grandmother, Jake Moffatt asked the chatbot on Air Canada’s website about bereavement fares. The assistant told him he could buy his ticket at the normal price and claim the discount afterwards, “within 90 days of the date your ticket was issued”. It wasn’t true: the airline’s policy did not allow the discount to be applied retroactively.

Moffatt complained. Air Canada refused the refund and offered him a CA$200 voucher instead. The case ended up before the Civil Resolution Tribunal of British Columbia, which ruled on 14 February.

“A separate legal entity”

The most striking thing is not the ruling but the defence. According to the decision, Air Canada argued that the chatbot was, in effect, a separate legal entity responsible for its own actions. The tribunal member, Christopher Rivers, called this “a remarkable submission” and replied with a sentence that ought to be framed on the wall of any department deploying a virtual assistant:

It should be obvious to Air Canada that it is responsible for all the information on its website. It makes no difference whether the information comes from a static page or a chatbot.

The tribunal found negligent misrepresentation and awarded a total of CA$812.02 in damages, interest and fees, according to The Guardian.

Why it matters outside Canada

This is a small-claims tribunal in another country, so it sets no precedent here. But it captures an intuition every customer shares and which, in Europe, fits the spirit of consumer protection law: if a company puts a tool to work speaking on its behalf, what the tool says, the company says.

What worries me is the opposite temptation, which I’ve seen more than once: treating the chatbot as an experimental “extra” whose mistakes are forgiven because “it’s AI”. A customer who gets an answer on a corporate website doesn’t care what technology sits behind it.

Four decisions before putting an assistant in front of the public

1. Limit what it talks about. An assistant that only answers questions about opening hours, deliveries or bookings gets things wrong less often than one that holds forth on commercial policies riddled with exceptions.
2. Answer from sources, not from the model’s memory. Answers about prices, deadlines or terms must come from official, up-to-date documents, with a link to them.
3. Know when to stay quiet. When in doubt about refunds, discounts or complaints, the right move is to hand over to a person. “I don’t know, let me put you through to the team” is not a failure; it’s design.
4. Own the mistake when it happens. If the assistant gets it wrong, the company is answerable. Better to have decided that in advance than to defend yourself with an argument that no tribunal, and no customer, is going to buy.

Conversational AI can greatly improve customer service. But the trust it generates is borrowed: the brand lends it. And it is the brand that pays it back when something goes wrong.