It’s Kentucky Derby week — the most exciting two minutes in sports.
What always gets me is the math. Months of breeding, training, conditioning, and strategy. Millions of dollars. Constant commentary and expert speculation. Then the gates open, and in roughly 120 seconds, you find out if any of it was real.
It’s a useful lens for what’s happening in AI right now.
Every organization I talk to has something in motion. Pilots, proof-of-concepts, working groups, use cases. The language is confident. The decks are polished. The demos are impressive. But when I ask “what’s actually in production?” — the room gets quieter.
The gap isn’t ideas. It’s never been ideas. The gap is execution. What makes it across the finish line. What still works once you’re past the demo and into the real world, with real users, real edge cases, and real scrutiny.
At the Derby, horses don’t lose because the training plan was wrong. They lose because race day is different from the track. The footing is different. The field is different. The pressure is different. The horses that win aren’t always the fastest in practice — they’re the ones that run their race when it counts.
AI projects fail for the same reason. The controlled environment of a pilot doesn’t prepare you for production. The demo that wowed the executive team doesn’t survive contact with the actual workflow. The use case that worked on clean sample data hits a wall when real data shows up.
I’ve spent 30 years in technology — a big chunk of it in regulated industries where the stakes for getting it wrong are real. And what I’ve learned is that the organizations that successfully scale AI aren’t doing anything dramatically different in the ideation phase. They’re doing something different in the execution phase: they treat deployment like a race day, not a rehearsal.
That means:
• Defining done before you start (not “working in the lab” — working in production, with users, under real conditions)
• Building for the edge cases, not the happy path
• Treating the first production deployment as mile one, not the finish line
The horses in the Derby didn’t get to race day by accident. Neither do AI projects that actually ship.
Unfiltered Labs exists because I got tired of watching good ideas stall before the gate. The only thing I care about is what crosses the finish line.