AI Is Killing Two of Three Product Manager Roles,-Keynote Argues

Some of the industry's loudest voices now declare the product role dead, while others call it the single most important role going forward. A keynote prepared for the 2026 Product Leader Summit argues both camps are describing different jobs — and AI is about to eliminate two of the three.

The core claim: the operating model a company uses to create products drives the roles on its teams, and the industry has long run two very different models side by side. The confusion barely touches designers and engineers. It hits product managers hard, because three distinct roles share one title.

What are the three product roles?

In the project model, companies run roadmaps of prioritized features and projects, each a hoped-for solution to an unstated problem. Success means shipping the prescribed solution. Two roles live here:

• Feature team product manager — marshals and coordinates resources to design, build, test, and ship a given solution. In practice this is a project-management job, with the PM writing docs one day, doing QA the next, and committing simple front-end code after that.
• Product owner — manages the delivery backlog in Agile (Scrum, Kanban) or SAFe environments, coordinating with engineers before, during, and after sprints.

In the product model, leaders assign teams problems to solve plus a desired outcome. Success means achieving that outcome. This produces the third role:

• Empowered team product manager — helps discover a solution that solves the problem and hits the outcome, bringing deep customer, market, and domain knowledge plus broad business viability knowledge across sales, marketing, finance, monetization, legal, compliance, security, and privacy.

The takeaway: one title, three jobs with very different responsibilities and skills, and most of the product community behaves as if only their version exists.

Which roles does AI eliminate?

Companies are racing to use agentic AI to automate delivery, building software factories where agents handle the work previously done by developers, QA, release managers, and feature team PMs, with engineers supervising the agents.

For product owners, the verdict is blunt: with AI automating and accelerating engineering workflows, and much of the sprint process obviated, the role is already being phased out. The keynote author has urged product owners to upskill as fast as possible for the past couple of years and expects the dedicated role to resolve itself.

For feature team PMs, the future is least clear. Some will upskill toward empowered roles if their companies adopt the product model. But most companies, driven by a desire for predictability, will likely never make that move. With automation eating so many feature team responsibilities, it is unclear how much of the job will remain — perhaps some oversight beyond what engineers and their agents provide.

Empowered team PMs face the opposite trajectory. Once companies realize delivery is no longer the bottleneck, the real question becomes what to build. That is the craft of product discovery: building to learn, not building to earn. Discovery means testing prototypes against four risks — value and usability with users, viability with stakeholders, feasibility with engineers — using qualitative and quantitative techniques. This demand should keep rising for at least the next several years.

The distinction matters most for probabilistic products, where unit economics, legal liability, safety, privacy, compliance, and ethics risks are considerable, and where the empowered PM necessarily sits at the center of the eval process — the mechanism for judging whether a result counts as good or even acceptable.

Why delivery automates and discovery does not

Product delivery consists of essential workflows, which agentic AI automates well. Product discovery is defined far less by workflows and far more by product sense and judgment, which resist automation. AI still contributes: it helps assess market dynamics, competitive landscapes, and customer behavior, and an AI Product Coach can help develop product sense and support discovery activities.

Shreyas Doshi recently framed the hiring signal: "Product Managers who've historically viewed their role as 'my main job is to execute' are finding it way harder to get hired at top companies than those who've viewed their role as 'my main job is to make the product successful'."

Execution was never the PM's main job, the keynote argues — enabling product success was. Automating the software factory is worthy work, but if a PM's higher-order responsibility is achieving outcomes, the focus belongs on discovering a solution worth building.

As agentic AI compresses delivery further, expect the market to price discovery judgment ever higher — and execution-only PMs to find their title, and their leverage, disappearing.