Notes from the field, kept here.

Last week my AI flagged one of my kitchen team members as an overtime risk.

This week I asked it a different question:

“Who on my team deserves recognition that hasn’t received it?”

Same person. Same hours. Same data.

She comes in early, again and again, to cover the prep station for a teammate who has been out for health reasons. She has never asked for credit. She also has not touched our team chat in 30 days, so none of this showed up anywhere I normally look.

I wrote her a thank-you note in Haitian Creole, her first language, and gave it to her myself.

Then I went looking for the teammate she was covering for. We do not share a language. I got a hug and a smile anyway. Some things do not need translating.

An overtime risk and an unsung hero can be the same person. It depends on what you ask.

The AI did not change. The question did.

That is AI for Significance prompt #04. Steal it. Ask it about your team this week, then go do the human part.

Scheduling hours in a Chick-fil-A is a 2,000-piece puzzle where the pieces change every week.

Here is what that puzzle looks like with 90 people on the team:

The minimum viable crew to open. The lunch rush that triples it. The afternoon taper, the dinner build, the close. School schedules. Minor work laws. Time-off requests. The kid who wants 10 hours a week next to the grinder who would take 70 if I let him. The single mom feeding a family next to the teenager whose parents made him get a job.

I have run hundreds of AI simulations to optimize it. Real progress. But here is what I keep finding:

AI treats humans like numbers on a spreadsheet. Mathematically correct. Personally blind.

So we keep humans in the loop. AI proposes. A leader who knows the Tuesday closer just started night classes decides.

That is the line I will not let AI cross: it can do the math on my people’s hours. It does not get to do the judgment on my people’s lives.

I hope to crack the code eventually. Until then, the puzzle gets solved by a machine and a human, together.

What is the 2,000-piece puzzle in your business?

The most capable AI model on the market spent three weeks under a U.S. government ban. Today it came back.

Quick version: Anthropic shipped Claude Fable 5 on June 9. Three days later, Commerce stepped in after researchers showed the model could be pushed past its guardrails to find software vulnerabilities. Three weeks of review later, the restrictions lifted. Today it’s back.

Here’s what I told my managers, and what I’d tell any non-technical owner watching from the sidelines:

1. The headline is scary. The process is the good news. A capability got flagged, everything paused, it got reviewed, it came back. That’s what maturity looks like in a young industry.

2. Your business doesn’t run on launch-day drama. It runs on boring, reliable systems. I run ten of them.

3. Power without evaluation is just risk. Same as hiring: the impressive resume still has to survive a Friday rush.

I use Claude tools in my restaurant every day, so Fable 5 enters my evaluation queue this week. When I have receipts, numbers not vibes, I’ll share them here.

What would you actually do with a smarter model in your business?

I run 10 AI systems inside my Chick-fil-A.

I am not an engineer. I taught systems engineering at West Point, but I have never written production code.

What they do: Forecast next week’s sales so we prep the right amount of chicken. Track inventory so my managers stop counting boxes. Screen applicants so we interview the right people faster. Build schedules that used to eat a manager’s entire afternoon.

Here’s what surprised me. AI didn’t replace a single person. It exposed something instead: which of my managers were leading people, and which were managing tasks and calling it leadership.

The busywork was hiding the difference. AI removed the hiding place.

If you handed one workflow to AI this quarter, which one would it be?