Air Canada's AI Chatbot Cost Them in Court. Here's What Every Business Leader Must Know

Air Canada's chatbot gave a passenger wrong information about bereavement fares. The customer followed it, booked the flight, and applied for a refund. Air Canada rejected it.In court, Air Canada's defense: "The chatbot is a separate legal entity."The tribunal rejected it. Air Canada lost.You are liable for what your AI says — even if you can't explain how it arrived at that answer.In this episode, I sit down with Malcolm Hawker (Chief Data Officer at Profisee, former Gartner analyst, 1,500+ CDO conversations) to unpack what actually breaks when companies deploy AI on ungoverned data.What we cover:Why 95% of AI pilots fail (governance breaks first, not technology)The Rule of 10: Fix data after = 10× cost, use bad data in decisions = 100× costHow Lexmark generated $2M in additional revenue from answering one question: "How many copiers do we have?"The "Turn It Off" moment: Why a CEO rejected accurate data (it broke sales compensation)Jevons Paradox: Why good governance creates MORE demand, not lessThe 15% vs 85% divide: What separates companies who ship customized AI from those stuck on "best effort"IBM's guardrail paradox: Using AI to police AI's biasWhy legacy frameworks (DAMA wheel) consume 2-3 years with zero ROIWhat Malcolm's seeing in production: Explainability becomes the only defenseGuest Background:Malcolm Hawker is Chief Data Officer at Profisee and author of The Chief Data Officer's Playbook and The Data Hero Playbook. As a former Gartner analyst, he's had over 1,500 conversations with Chief Data Officers and seen what works — and what fails — in production AI deployments.For Leaders Who Need to Decide:If you're a CDO, CTO, or AI product leader deploying AI this quarter, Malcolm offers field-tested frameworks you can test immediately:Don't start with frameworks — start with outcomesHire a value engineer (quantify governance in CFO language)Go outcome by outcome (not "fix it all" strategies)Recognize that good governance unleashes demand (Jevons Paradox)Ask: Can you explain your model's output to a judge?This isn't theory. These are patterns from 1,500+ CDO conversations and real production deployments.Connect with Malcolm:LinkedIn: https://www.linkedin.com/in/malcolmhawkerHost:Subrata Kar studies patterns from builders who scale — enterprise systems, AI platforms, and startups — and extracts actionable insights leaders can apply immediately.LinkedIn: https://www.linkedin.com/in/subrotoNewsletter: https://substack.com/@subratakar