Platform / technical depth
Data and integration was always the product. AI just made it obvious.
The experience a product team builds is a proxy. The actual product is what persists when the experience is removed: the data and the quality of its integrations.
At AWS Health, we made this pivot before the AI wave forced it. The product had three separate experiences: a console, an email channel, and an API. Each had its own path, its own surface, its own definition of how customers were supposed to engage. The problem was customers weren’t following any of them.
The shift wasn’t to build a better console or a better email. It was to make a different commitment: the same data, every surface. Not three separate experiences with different data, one integration layer that any surface could draw from. Then a maturity framework that gave customers an incremental path to embed that data into their own workflows: here’s where you are, here’s what’s next. The customers build the path. We built the on-ramp.
The work that followed wasn’t feature work. It was friction removal, working backwards from where customers got stuck, not forwards from what we could build. Coverage grew to a majority of customers, organizations, and revenue not because we added capabilities, but because we removed the barriers to using the ones we already had.
That would have been enough in 2023. With agents, it’s the minimum bar.
An agent integrating your product needs consistent data across every touchpoint. If your API returns different fields than your webhook payload, the agent confidently returns the wrong answer. If your integration model has rough edges (inconsistent schema, undocumented behavior, friction in the auth flow) the agent can’t bridge them the way a human operator could. Rough edges that a determined customer would work around become hard stops for automation.
The inverse is also true. A clean interface, consistent data, a clear model for how to embed your product into a workflow, and an agent can integrate your product into a customer’s stack without a sales cycle, without an onboarding call. Adoption becomes programmatic.
Products process inputs into outputs: data in, decision or action out. The UI helped humans facilitate that transformation. It was never the cause of it. Agents don’t need the facilitation layer. They work directly with the data. The UI isn’t being improved. It’s being bypassed.
The PM who built for control of the journey is watching AI route around it. The PM who built the clean interface is watching agents adopt their product automatically.
Data and integration was always the product. AI just made that impossible to ignore.