This week’s Chatbots Behaving Badly episode is a debate disguised as a bar conversation.
On one side is the shiny future: bots that interview guests and recommend cocktails, agents that connect to inventory and operations, and an emerging wave of “smart hospitality” systems that promise personalization at scale.
On the other side is Harry, a London bartender who refuses to be called a mixologist and sees himself as part of a lineage that runs straight through Harry MacElhone, the founder of Harry’s New York Bar in Paris. Harry’s argument is simple: AI can copy recipes, but it can’t do the job. It can’t read the room, sense hesitation, slow someone down without humiliating them, or keep a night from tipping over the edge.
That clash becomes most interesting when we get to the feature everyone thinks is “responsible” until they think about it for ten more seconds: measuring intoxication. Because once your system measures impairment and still serves, you’ve created evidence and moved into duty-of-care decisioning. The future doesn’t just get automated. It gets documented.
We also talk about mood inference, why camera-based “emotion AI” is a fast track from hospitality to surveillance, and what a sane version of the future actually looks like: explicit choices, consent-based personalization, minimal data, and humans staying in the loop where judgment matters.
If you’ve ever heard someone say “we can just personalize it,” you’re going to enjoy this episode.
