When AI Makes Promises Your Company Can’t Keep

How often do you visit a company’s website to ask about a discount or shopping assistance? Nowadays, many websites use friendly chatbots to explain policies, qualification requirements, and other instructions to answer simple questions.

So imagine that you follow those instructions…only for the company to tell you the chatbot was wrong all along.

Who’s responsible for the mistake?

That question became the center of a real dispute involving Air Canada, and it offers an important lesson for anyone using AI in marketing, customer service, or business communications.

AI can make customer communication easier. It can also confidently make promises that no real human authorized.

What Happened at Air Canada?

In November 2022, a customer named Jake Moffat needed to travel after a family member's death. While researching flights, Moffatt asked Air Canada’s website chatbot about bereavement fares, which offer reduced prices under certain circumstances.

The chatbot explained that Moffatt could purchase a ticket and apply for the reduced fare afterward, within 90 days. Unfortunately, none of that was true.

Air Canada’s actual policy did not allow customers to request the discount after completing their travel.

Moffatt followed the chatbot’s instructions, however, purchased the tickets, and later requested a partial refund. Air Canada refused.

The disagreement eventually reached British Columbia’s Civil Resolution Tribunal, and in February 2024, the tribunal ruled against Air Canada, finding that the airline had failed to take reasonable care to ensure its chatbot provided accurate information. As a result, Air Canada had to pay approximately C$812 in damages, interest, and fees.

The lesson? A company cannot necessarily escape responsibility for incorrect information simply because AI delivered it. If you own the chatbot, you’re responsible for verifying the information it produces.

Your AI Assistant Represents You

Think about how businesses use AI today. Chatbots answer customer questions, marketing platforms generate personalized emails. AI tools recommend products, explain services, and create promotional content.

Those capabilities can be extremely useful. The problem arises when AI starts communicating information that customers (reasonably) believe comes directly from the company.

Imagine a chatbot incorrectly telling someone:

• A product includes a three-year warranty.

• A subscription can be canceled without fees.

• A particular service guarantees specific results.

• A customer qualifies for a discount that does not exist.

Even if nobody intended to make those promises, customers may act on them. That creates financial, reputational, and potentially legal risks for everyone.

Keep AI Within Its Limits

AI can still be incredibly valuable for marketing and customer engagement. The key is establishing boundaries.

• Use verified company information when building customer-facing AI tools.

• Review important claims about pricing, warranties, refunds, and product capabilities.

• Require human approval for sensitive communication and make sure customers can reach a real person when necessary.

• Avoid uploading confidential customer information into unapproved AI services.

Most importantly, regularly test what your AI tools actually tell people. A chatbot that worked correctly last month might still provide an incorrect answer tomorrow.

Conclusion

The Air Canada case shows that trusting information from a company’s integrated AI can lead customers down the wrong path, costing you—and them—valuable time and money.

That same problem can affect marketing emails, advertisements, chatbots, and automated customer communications. While AI can help you reach more people, answer questions faster, and create content in seconds, it also has its risks and drawbacks.

Remember, customers may not distinguish between information they get from your coworkers and your company’s AI-generated answer. If a chatbot speaks on your company’s behalf, then you still need to make sure it knows what it’s talking about!