I build strong product teams in the age of AI.

Two threads run through fifteen years of this. The first is technical: take raw, messy data — card transactions, healthcare claims, endpoint telemetry, model output — and turn it into something a business will pay for and a customer will trust. That one runs from security software at LANDESK through Adobe, MX, Mastercard, and now Ivanti.

The second is the one I've come to care about more. I've built the product function where there wasn't one — as CPO at UHIN, standing up the practice and the operating cadence from zero. I've turned an engineering-led org into a product-led one by putting PM, design, and engineering into customer conversations together instead of handing them a backlog. And I've coached product managers through the gap between running a backlog well and owning a strategy, including one who made Director on the strength of the discovery skills he didn't have when we started.

What I lead now is platform work, and platform work is its own discipline. A shared AI capability isn't finished when it works. It's finished when another team can adopt it faster than they could rebuild it — which makes developer experience, golden paths, and self-service onboarding part of the product rather than something you get to afterward. I'm hands-on about it: I'd rather prototype with agentic tooling than describe what I'd like someone to build, and I'd rather let eval evidence settle an argument than a well-made slide. A lot of my recent thinking is about what that changes for how teams work — it's in the writing.

What AI changes — and what it doesn't

It doesn't change what a product team is for. Using AI to aggregate requests, generate roadmaps, and write PRDs faster doesn't move a team to the product model. It runs the project model at a higher frame rate — Cagan calls that product management theater, and the label is deserved. Outcomes over output survives the technology change completely intact.

It does change the cost of being wrong. When discovery evidence takes an afternoon instead of a quarter, the honest move is to raise how often a team is willing to be wrong in private, not to ship faster in public. Most of what I've built at the platform layer — evals, traces, an agent framework teams don't each rebuild — exists to shorten that loop for everyone else.

It changes where the coaching bottleneck is. Cagan's recent argument is that a foundation model can serve as a personal product coach, and for craft — sharper problem framing, better eval design, an honest critique of a strategy doc — I think he's right, and it's genuinely democratizing. What a model can't supply is strategic context or the organizational cover a PM needs to make a real decision. Nor does it make a team ready to be empowered; empowerment is an outcome of coaching, not a substitute for it. That part was always the leader's job and still is.