I built myself an operating model and then it told me I was wrong

Six months into my current project, I could feel something was off. I had a methodology analysis, a working prototype, an AI behavior framework, a research plan with 25 participants identified, and a lot of momentum. But I kept noticing that the momentum was in analysis, not in learning. I hadn’t talked to a single person from the target population.

I knew what the problem was, I just didn’t have a structure to act on it. So I built one.

A four-phase innovation loop (Sense, Shape, Test, Scale) with time-boxes and go/pivot/kill criteria at each gate. None of these ideas are original on their own. The loop draws from IDEO’s design thinking, lean startup, dual-track agile, things I’ve used at previous jobs and patterns I’ve picked up from Teresa Torres, Eric Ries, and others over the years. What I needed was to combine them into something specific enough to hold me accountable on this project.

When I ran my own project through it, it confirmed what I already suspected. I’d been going deep where the doors were open (the research archive was accessible, the academic literature was available, Claude Code made analysis incredibly productive) and deferring action where the friction was (approval processes, recruitment logistics, the general discomfort of putting unfinished work in front of real people). The framework gave me a way to name that clearly and course-correct with a plan, not just a feeling.

The correction was fast once I had the structure. Within a few weeks I went from zero user contact to a market test, a pivot decision backed by actual behavior data, and a build plan that came from what people did, not what I assumed they’d do.

I’m still refining the framework, and I don’t think it’s something I’ll ever stop adjusting. But building a tool to check my own work, and then actually using it honestly, is the practice I keep coming back to.