Two years ago I managed products. Today I orchestrate builds across humans and AI agents — simultaneously. At Machinify, that means turning five acquired healthcare companies into one platform. $200B+ in claims. 160 million lives. A workflow where agents draft, humans judge, agents iterate, and humans ship.
At Machinify — unifying five acquired healthcare payments companies into one AI-native platform. The work is designing systems where AI agents are first-class participants. Not prompted tools. Not hypothetical teammates. Agents that get assigned work, take action, flag problems, and learn from outcomes.
It's the most complex thing I've ever built. Also the most fun.
Thinking about: what happens to organizations when the PM-to-engineer ratio inverts. Writing about it at The Big Picture. Building with it every weekday.
I keep getting hired to do the same thing in different industries. Find the mess. See the system hiding inside it. Build the infrastructure layer that makes everything else possible. The technology changes. The pattern doesn't.
Unifying five acquired healthcare payments companies into one AI-native platform. Agents get assigned claims work, take action, flag problems, learn from outcomes.
Pioneered server-driven UI and intelligent configuration. Called the AI shift before it was fashionable, built the architecture to survive it.
Told the company to stop shipping features and build the platform. Was right inside six months. Then a pandemic arrived and we quadrupled through it.
Turned an analytics company into an AI-powered healthcare SaaS platform. First place the infrastructure thesis really clicked.
Twelve years writing code. Semiconductors, distributed databases, enterprise systems. The foundation every product decision since has stood on.
The organizations that figure out human-agent collaboration first won't just have an advantage — they'll have a compounding one. Most organizations are not remotely ready. Most of their tooling isn't either.
Twelve years writing code before a decade in product. That combination used to be a bonus. It's now the baseline.
Which means the cost of building the wrong thing is the only cost that matters. That changes what product people are for. Most job descriptions haven't caught up.
Unusual combinations of knowledge — engineering + product + healthcare + AI — are the new moat. Specialists with one depth are the ones I'm worried about.
They make everything else possible without calling attention to themselves. If your platform has a brand, it's probably not a platform yet.
Prototypes replace slide decks. "Let me show you" beats "let me explain to you." A working thing ends the meeting.
People who still think the job is writing documents are being automated away. People who become builder-orchestrators are doing work that didn't exist two years ago. Pick which one you're becoming.
"Phil possesses a remarkable foresight — particularly evident in his early and accurate conviction regarding the transformative potential of AI."
He called the AI shift before most of us could spell agent. We built the platform around that thesis and it was right.
"Deep technical understanding with sharp product instincts. Rare. Critical as we pushed the boundaries of what an AI-native healthcare platform can deliver."
Phil is the PM who can read the PR and redline the architecture in the same sitting. It changes the pace of everything.
"Phil immediately saw the opportunity to propel our company to a position of digital leadership by building an extensible platform — rather than endlessly rolling out one feature after another."
He walked in, said "stop shipping features, build the platform," and was right inside six months.
"With his leadership, we quadrupled the revenue we bring in through our portal. Phil is a world-class Product Director."
Something broke in the last six months. Not broke in a bad way — broke like a dam breaks. Everything that was building up behind it is now rushing through.
On assigning work to something that doesn't have a Slack avatar, doesn't go home, and still manages to surprise you.
Five companies, 160 million lives, and one architecture that has to disappear into the floorboards before anyone trusts it.
The working thing ends the meeting. A short argument for showing before telling, in a field that is addicted to telling.
I'm always up for a conversation with other operators in the collapse. Advisory calls. Quiet peer reviews of agent-native architectures. The occasional "is this crazy or is it the future" dinner.