The sharpest thing I read this week was not about product. It was about marketing. Chris Koehler, the CMO of Twilio, drew a line most leaders are still smudging: there is a difference between using AI and being AI-first. Using AI means you run the process you already have faster. Being AI-first means you rebuild the process assuming the machine was in the room on day one. Swap the word marketing for the word product and the argument does not weaken. It gets louder.
Most teams are optimizing a conveyor belt
Here is the uncomfortable read on where most product orgs are. They bought the tools, expensed the seats, and pointed all of it at the workflow they already had. Faster tickets, faster PRDs, faster teardowns. That is real, and it is worth something, but it is a conveyor belt running at higher RPM, not transformation. Being AI-first is asking whether the spec, the handoff, the sprint ritual, and the three-team relay race should exist at all. One is sanding the edges of 2019. The other is building for 2027.
The capacity gap is the real story
The most quietly radical moment in Koehler's piece is when he polls a room of CMOs on how many cut headcount as AI spread. The answer was none. They all carry a capacity gap, more they want to build than people to build it. Product has lived inside that gap longer than anyone. Every roadmap is a graveyard: for each thing that ships, ten good ideas lost a prioritization fight and twenty never made the doc. AI does not shrink the product team. It finally lets the team clear a decade of debt.
Specialists become the machine, taste becomes the moat
Koehler's prediction is that teams move from roughly 90 to 95 percent specialists toward horizontal generalists, with AI as the specialist inside each lane. The model is a film crew, not a production line. The PM whose value was coordinating specialists, routing tickets, and chasing status is being automated. What compounds is the person who holds the vision and moves across design, code, data, and go-to-market, using AI for the deep work in each lane.
And once building is cheap, the scarce input is taste. Features used to be expensive, so shipping was the constraint. Now the constraint is judgment: deciding what not to build, and having the nerve to subtract.
The move
Stop asking how AI makes your current process faster. That question has a ceiling and you are already near it. Ask the harder one: if you rebuilt your product organization from scratch this quarter, knowing what these tools can do, what would you never reassemble? Pick one ritual this week, the standup, the PRD template, the quarterly planning offsite, and ask whether it would survive a clean rebuild. Start there.