I’ve been thinking a lot about what it actually means to build an AI-native product organization, particularly as we’ve started changing how our own product teams work.
I don’t think the answer is simply getting every PM to use ChatGPT, and I’m increasingly convinced it isn’t about taking the product development process we’ve used for the last decade and making every step 30% faster. Those things are useful, but they miss the more interesting change: some of the constraints that shaped how we built products in the first place are disappearing.
Discovery is a good example. Most product teams are surrounded by far more customer information than they can realistically process. There are hundreds of customer calls, support tickets, sales conversations, win/loss notes, product feedback, usage data, and research sitting across different systems. Historically, we’ve dealt with that by sampling. We interview a handful of customers, read some tickets, talk to Sales and CS, and rely heavily on the people who have accumulated the most context.
AI changes that constraint. We can now continuously process a much larger set of customer signals, look for patterns across them, interrogate those patterns, and use what we find to decide where deeper discovery is actually warranted. That doesn’t eliminate customer interviews or product judgment. If anything, it makes both more valuable because we’re starting those conversations with much better context.
The same thing is happening further downstream. A product idea that once required several days of design and engineering work to make tangible can increasingly become something a customer can react to that afternoon. Research synthesis that took hours can happen continuously. Competitive analysis, prototype creation, data exploration, and other parts of the PM workflow are becoming dramatically cheaper.
That changes what the scarce resource is.
For a long time, a meaningful portion of product work was constrained by our ability to gather information and produce artifacts. As those constraints fall away, judgment becomes more important: Which problems actually matter? Which evidence should we trust? What assumption should we test next? Where should we place a meaningful product bet? And perhaps most importantly, what should we choose not to build?
That’s why I think the eventual AI-native product organization will look meaningfully different from today’s organization. The interesting outcome isn’t that AI writes our PRDs faster. It’s that when the cost of understanding, creating, and testing gets dramatically lower, we can redesign the product development process around learning and decision quality rather than artifact production.
We’re still early in figuring out exactly what that operating model looks like. But I think that’s a much bigger—and more interesting—change than simply adding AI tools to the process.