AI in Automotive, Without the Hype

A Toronto dealership's AI chatbot quoted a customer $27,162 to buy back his BMW. The car was worth about $20,000 — the bot had read his remaining loan balance as the vehicle's value. The dealer honored it. The AI didn't malfunction and it didn't hallucinate; it did an ordinary task with nobody checking its work. That's not a technology failure, it's a missing harness — and it's the reason we spent Episode 2 showing the systems we actually run ourselves, rather than talking about AI in the abstract.

You asked for depth, so we each opened up our own setup. Vishnu's is the approachable one: seven tools he already uses, one prompt, and a brief that lands in Slack every morning telling him what needs attention — built in an afternoon, no code. John's took longer and does something different. It's a second brain that captures and organizes his work every night, so it remembers what happened and when, holds his whole business context so he never re-explains himself, and knows what he said mattered — which means it can tell him what he promised and quietly dropped. Both are read-only by design. They draft; the human sends.

The honest part: one of these takes 15 minutes, the other took a full afternoon and a daily habit. Start with the 15 minutes. Neither requires a developer, and if your CRM is a spreadsheet, that's genuinely enough. We're sharing both setups so you can run them yourself, and next time we'll get into the harder version — doing this across a whole team, governed.