Holden Richardson
I build production AI systems inside real companies — not demos, not slideware. The work that taught me this was hands-on: I implemented agentic AI inside a 100-person West Michigan wire harness manufacturer, wiring agents into the document-heavy, rules-based processes that eat skilled people’s time.
The discipline is second-generation. My father was a hands-on manufacturing operations consultant before he owned a manufacturer — he moved machines, changed processes, standardized work, and managed vendors. He re-engineered the physical operation. I apply the same discipline to the information side: how knowledge work flows, where it stalls, and what a well-set-up agent can carry.
Everything gets measured in hours and dollars, not hype. Every published number carries a source, or there is no number.
That principle is why every install is vendor-neutral. The canonical skills are plain markdown any agent can execute — no provider lock-in, ever. When a better model shows up, you switch to it and keep everything you built. The system is yours.
