Medium: How AI is changing product team structure and decision-making

Joca Torres is a product management consultant with over 30 years of experience working with product teams. In this February 2026 piece on Medium, he gathers observations from teams that are actively using AI coding and prototyping tools and examines how those tools are changing team dynamics — not at the level of individual productivity but at the structural level of where work gets blocked.

The core observation

When engineering execution becomes faster, the bottleneck moves upstream. Torres documents anonymous cases where engineering teams reported moving faster than product teams could make decisions. This reverses the more common dynamic in which engineers wait on product to clarify requirements or prioritize backlog. The shift is not hypothetical — Torres presents it as something teams are currently experiencing rather than predicting.

What follows from faster execution

Several structural questions open up once implementation speed is no longer the primary constraint. The traditional staffing ratio of one PM to five or nine engineers was built around the assumption that engineering time was expensive and product direction was relatively cheap to provide. When implementation cost falls, that ratio loses its original justification. Torres frames this not as a recommendation to restructure teams but as an organizational question that some companies are beginning to confront.

He also notes an effect on knowledge distribution within teams. AI tools that help document complex codebases and summarize technical context are reducing the dependence on specific senior engineers as sole keepers of institutional knowledge. When that knowledge becomes more accessible through tooling, it changes how PMs communicate with engineering — less time spent building context, more time spent on direction.

Faster output also raises the need for stronger senior oversight. When a team can generate more features or iterations in a given time, the pressure on architectural review and quality assurance increases rather than decreases.

What the article does and does not offer

Torres writes from consulting experience, so the examples are anonymized and cannot be verified independently. He does not propose a restructuring framework or a recommended ratio; the article is observational rather than prescriptive. That is also its limitation — readers looking for a concrete model to bring to their organization will not find one here.

The value is in the diagnosis. For PMs at companies actively adopting AI coding tools, this piece offers a clear-eyed description of what is actually shifting organizationally, beyond the individual productivity gains that most AI-in-PM writing focuses on.