Lightening Case Study: Kaiser Permanente + Google GenCast

I tried something new this week. No guest, no interview, just me thinking out loud about a talk I'm building on how companies actually get value from AI.My complaint is simple. Most of us use AI to do the same work faster, and that shortcut has a ceiling. It's why the ROI never shows up the way it's promised.So I went looking at how scientists use AI instead. Google DeepMind's weather model, GenCast, never got taught a single rule of physics. Researchers just fed it decades of real weather and let it find the pattern itself. It beat the gold standard forecast on 97 percent of tests.Kaiser Permanente ran the same play on patients. Their Advance Alert Monitor catches the small signals someone's about to crash before the alarm goes off, not after, and gets a nurse there while there's still time.Same move both times. Feed it everything, name the outcome, let it work backward. Here's the rule I'm building my talk around. Give AI the destination, not the directions.I'm not done. More case studies coming, and I want you along for the ride.About The Job Sh*t ShowThe Job Sh*t Show is an investigation into how finding, keeping, and doing work is changing faster than anyone is admitting — and what that actually looks like for real people on the ground. The Sh*t List is a paid Substack for leaders who want to stay ahead of what AI is doing to the workplace — and actually do something about it. Every other week, subscribers get a new AI Manager Skill: one practical, field-tested move for leading your team better in an AI world.