I have been through two big shifts in how product gets done. Waterfall to agile took most companies a decade. The second one, human-only teams to human-plus-AI, is happening now and going faster. The operating model is how I keep the parts of the old practice that still work while rewiring the parts AI broke.
Keep the foundation
The pre-AI model was correct. Marty Cagan's empowered teams, Teresa Torres's continuous discovery, the PM-designer-engineer trio. Its core principles carry straight forward: customer obsession as a daily practice, small teams with real authority over an outcome, and measuring whether customer behavior changed instead of whether you shipped the thing. The temptation with AI is to throw everything out. Do not. These get more important, not less.
What AI breaks
The spec is dead. I have not written a PRD in over a year, not because I got lazy, but because I can build a working prototype faster than I can write a document describing it. The prototype is the spec. Customers react to a real thing, and the feedback is on a different level.
The trio becomes a quartet or a duo. A designer can now prototype without waiting for engineering. A PM can run analysis that used to need a data scientist. Roles are not disappearing, but the boundaries are blurring, and each person shows up to the sync with more done.
Discovery compresses. The old cycle was four weeks: interview, synthesize, prototype, decide. Now agents monitor behavior over the weekend, I review synthesized insights over coffee, we generate prototype variations the same day, test with customers, and decide. A month becomes a week. PM is a learning-speed game, and AI compressed the loop hard.
Execution overhead shrinks. I used to spend about 40% of my time writing tickets, updating dashboards, building status reports, and grooming backlogs. Agents draft release notes, watch dashboards, and write first-draft status updates now. That time goes back into discovery and judgment. AI does not replace the PM. It kills the parts of the job that were never the real job.
The new weekly rhythm
Monday I start by reviewing AI-synthesized insights instead of pulling reports myself. Wednesday produces working, clickable prototypes instead of wireframes. Friday is the big one: it went from "plan what to build" to "ship what you built," because Wednesday's prototypes are already close to production-ready. That is not a tweak. It is a different operating rhythm.
Pick one thing this week. Take a feature off your roadmap, skip the spec, describe it to an AI coding tool, and put the prototype in front of a customer. Keep the principles. Rewire the execution.