From execution to coordination: product and engineering

Over the last few months, much of the AI tooling I have been following in software has targeted code execution: tests, bugs, docs, security. Automating the technical lower layers. The new wave of agentic tools shifts towards finished knowledge work: analysis, reporting, planning, and all with minimal supervision. That’s automation moving from execution into coordination.

If you abstract the PM role, much of it is cognitive orchestration: understanding context, synthesizing inputs, defining problems, prioritizing, specifying intent, aligning stakeholders, validating outcomes. These are exactly the structured knowledge workflows agents are starting to absorb. So yes, parts of the Product domain are now in scope for automation, just like parts of engineering were until now.

The shift I see in engineering is toward system design, orchestration and problem framing. That is where I expect more of our value to sit as implementation becomes easier.

A similar shift is reaching product. What remains is the PM as a decision architect inside increasingly automated socio-technical systems. Which is also why product and engineering are converging: engineers move toward product definition as implementation friction drops, while PMs move toward system design as coordination becomes programmable.

Rich Holmes’ example of a product workflow using Claude Cowork prompted this reflection.

Here is another flow from the same source: Rich Holmes’ example with the Ideate Agent in Stitch