In February 2024 a Canadian tribunal ruled on a dispute about wrong advice from an airline's chatbot. According to an incident report, "The airline later argued the chatbot was a separate legal entity" (incidentdatabase.ai, report 3673). Daily Hive reported on 15 February 2024 that the tribunal did not accept that argument.
This is not legal advice, and the case was about a customer, not an employee. The practical lesson still carries over: if your bot says something, people will treat it as your company saying it.
Why it matters for a team of five or fifty
An internal bot that tells a colleague the wrong refund rule, leave policy or discount limit does not fail quietly. Someone acts on it, and your name is on the decision.
Four habits that lower the risk
- Answer only from facts a person approved. If there is no approved answer, do not guess.
- Show where each answer came from and when it was confirmed, so a reader can judge it.
- Say I do not know and send the question to a person. A clear refusal costs a few minutes; a wrong answer can cost a day.
- Keep a record of what the bot said and when, so a dispute can be settled from facts.
What a good refusal looks like
"I do not have an approved answer to that. I have asked the founder and you will get a reply here." It is short, it does not pretend, and it creates a task for a person.
Try it
AskMyStartup answers only from answers the founder approved, shows the source and date, and sends everything else to the founder. It is in private beta; join the waitlist from the buttons on this page.