In The Four Cs Before the Prompt I argued that context, conversation, and confirmation have to travel with every user story now, because an agent builds exactly what the issue says. The obvious objection: isn’t this the slow, document-heavy process AI teams are leaving behind?
Cat Wu, who leads product for Claude Code, wrote one of the best pieces this year on product management on the AI exponential. Her team plans in short sprints, replaced docs with demos, and treats every model release as a reason to revisit what they’ve built. In her words, you’re “building on ground that’s rising underneath you.”
Read closely, though, and what she describes is research at a different tempo.
Demos go to internal users, and only “the ones with real engagement get polished.” That’s a usability test. When the team noticed users building apps in one tool and copy-pasting instructions into another to test them, they turned it into a feature: “If users are hacking something together, that’s scaffolding you can build into the product.” That’s field observation. Her conclusion about the PM role is to “identify the handful of true non-negotiables and let the rest go.”
Taste, thought, and trust are those non-negotiables, and you only find yours by watching people use what you built.
So the cheap prototype doesn’t replace the conversation. It makes the conversation better, because you bring something real to the person instead of a hypothetical. But research has to change shape with the tempo. It can’t be a phase at the start anymore. When building was slow, a wrong direction revealed itself over months. Now a team with AI can build a complete, polished, wrong product in a week.
Speed doesn’t make skipping research cheaper. It makes it more expensive.
Every prototype is a question you take back to a real person. Every “it works” gets followed by “and do they trust it?”
The floor rose and the ceiling stayed where it was. Taste, thought, and trust are still up there, and the only way to reach them is still the slow part, done more often.
Sources
- Cat Wu, Product management on the AI exponential, Claude blog
Daniel Arevalo is a Product Architect and strategic product design leader. He helps product teams turn AI capability into trusted, adoptable workflows, with a background in industrial AI, IT/OT cybersecurity, and high-stakes enterprise software for critical infrastructure.