In 2024, Air Canada's chatbot gave a customer the wrong information about bereavement fares — and when he filed a claim, the airline tried to argue the chatbot was basically its own separate entity, not their responsibility. The tribunal said no. And that ruling is the whole episode: you don't get to outsource the judgment and keep the authority. Those two things travel together.
In this episode
Why most AI overuse starts with completely reasonable tradeoffs — "I'll fix it after," "just this once," "I'm sure it's fine" — and how those add up to a habit you didn't mean to form. The filmmaking adage "we'll fix it in post" and why it applies directly to what happens when you hand AI an underdeveloped prompt and hope for the best.
Two very different ways to ghostwrite a book — and which one most people are accidentally replicating with AI. The moment you stop developing ideas with AI and start approving ideas from AI, and why those are not the same job.
HumanPrint homework
Before you use AI for something that matters, write one sentence: "AI can help with ___, but I'm in charge of ___." Fill in both blanks. Then for one piece of work this week, mark every place AI helped and every place you made the final call. If the second list is thin, adjust the process. That's where your authority went.
