Community

translate Automatic translation · View original

From Product Owner to Product Architect: How AI Redefines the Role

From Product Owner to Product Architect: How AI Redefines the Role
The Product Owner was the voice of the customer on the team. Their job was to manage a backlog, prioritize user stories, and make decisions about what to build. For years, that was enough. Today, when a team can generate in hours what used to take weeks, that description of the role has become too small. AI has not eliminated the need for someone to decide what deserves to be built. On the contrary: it has made that need more urgent and more demanding than ever. There is an irony in what is happening to the product role in the age of AI. For years, the community debated whether the Product Owner had too much power or too little, and whether they should be more technical or more strategic. Then generative AI arrived, and that debate was overshadowed by a more concrete one: the role is not at risk of disappearing, but of becoming the point where everything is won or lost.

The Bottleneck Has Shifted


For decades, the main constraint in product development was engineering. Building software was expensive and slow, and the Product Owner managed that scarcity: they prioritized because not everything could be built, and refined because the team needed clarity to move forward. The design of the role made sense in that context.

What has changed is where the constraint lies. McKinsey, in its State of AI 2025, documented that 88% of organizations already use AI in at least one business function, but only 39% report real impact on business results. BCG reaches similar conclusions: most organizations are not generating material value despite their AI investments. The most consistent explanation does not point to the technology but to what surrounds it: teams produce features faster, but without the strategic direction that turns that speed into value for the user.

When building is cheaper, deciding well what to build becomes the differentiator. The constraint has shifted from engineering to judgment. And judgment about what deserves to be built, for whom, and why is, ultimately, the responsibility of the product role.

What the Original Design of the Role No Longer Covers


The Scrum Product Owner was conceived for a specific context: teams where engineering was the scarce resource, delivery cycles were two weeks long, and the backlog was the central coordination artifact. That design responded to the conditions of its time. In a context where AI agents can generate functional prototypes in hours and where much of the implementation work can be assisted or delegated, three classic PO responsibilities are being absorbed or assisted by tools: synthesizing user feedback, drafting first versions of user stories and acceptance criteria, and tracking backlog progress. What no tool can replace is the decision about which problem deserves to be solved, with what value criteria, and for which real user. That is the part that grows in weight and importance.

Marty Cagan (2024) argues that the risk lies not in adopting AI but in adopting it without rethinking how product teams are led and structured. Accelerating execution is not valuable in itself: it can just as easily speed up the delivery of the wrong things. That warning points directly to the product role as the place where it is decided whether speed translates into value or into faster waste.

The Product Architect: What Changes and What Stays


What stays is the essence of the role: representing the customer on the team, deciding what deserves to be built and in what order, and keeping the product vision as a reference. Teresa Torres has spent years arguing that the difference between product teams that generate value and those that do not lies in whether they do real discovery or merely validate solutions that someone already decided to build. In a context where building solutions has become much cheaper, that distinction becomes more critical, not less.

What changes are three concrete dimensions:

  • Shift toward continuous strategic validation. The classic PO validated at the end of the sprint, when the increment was already built. The Product Architect validates earlier: they design experiments to test hypotheses before the team builds, use the fast exploration lane to get user feedback in hours, and decide with data whether an idea deserves to move to production. McKinsey describes this change as a compression of the distance between discovery and delivery: AI makes it feasible to prototype and validate hypotheses at a speed that was previously unimaginable, multiplying the number of experiments a team can run. But that only generates value if there is strategic judgment behind what is tested and why.
  • Conscious management of the productivity paradox. When the team can generate features faster than the organization can assess whether they deserve to be built, the Product Architect needs to act as a deliberate brake. There is a difference between being a bottleneck (slowing things down without adding value) and being a point of strategic decision. This requires explicit criteria about what enters the backlog, what goes into the fast exploration lane, and what deserves the rigor of production.
  • Relationship with specifications as an artifact. User stories were designed to communicate intent between people. When part of the execution is carried out by an AI agent, that communication needs a different level of precision. The Product Architect does not have to write technical specifications; they have to ensure that the strategic intent behind each feature is clear enough for developers to translate it into a spec an AI can process. That requires a closer working relationship with the technical team, not a more distant one.

The Competencies That Gain Weight


Demand for product profiles with AI fluency has multiplied in recent years. Forrester found that job postings for "AI Product Manager" roles grew very significantly over a short period, which indicates that the market already recognizes this transformation even though the industry has not yet finished defining it.

This does not mean the Product Architect must become an AI technician. It means they need to understand enough about how it works to make good decisions with it and about it. Four competencies gain weight:

  • Data literacy: the ability to read user behavior data, product metrics, and experiment results to make evidence-based decisions.
  • Designing validation experiments: knowing how to formulate testable hypotheses, design the minimum test that confronts them, and read the results rigorously.
  • Managing the speed-value tension: judgment about when to go fast and when to slow down, when a prototype is sufficient and when production rigor is needed.
  • Communicating in processable language: not only conveying intent to people, but ensuring that intent can be translated into specifications that guide the work of agents.

What does not change (and it is worth underlining) is the need to truly understand users. AI tools can synthesize feedback at scale, analyze behavior patterns, and generate reports. They cannot replace the direct conversation that reveals the "why" behind behavior. In a context where the temptation to go faster is constant, spending time talking with real users is one of the few things that remains irreplaceably human.

A Transition, Not a Replacement


The Product Architect is not a new role that replaces the Product Owner. It is what the Product Owner becomes when the team's conditions demand it, and those conditions are more frequent today than before.

Not all teams need to make that transition now. Those that work with little or no AI assistance, in organizations where engineering remains the main constraint, can keep working well with the classic model. The question is not whether to change the title, but whether the role being performed responds to the team's real conditions.

What seems clear is that the organizations getting real value from AI are not the ones that have adopted the most tools, but the ones that have redesigned how product decisions are made. According to McKinsey State of AI 2025, the factors that distinguish organizations with the greatest impact include having defined processes to validate AI outputs and keeping user orientation as a central criterion. Both are responsibilities of the product role.

How is the product role evolving in your organization? What have you had to learn or unlearn since AI entered your team's workflow?

Bibliography


  • Battery Ventures (2025). The Role of a Product Manager in the Age of AI. battery.com
  • Cagan, M. (2024). Transformed: Moving to the Product Operating Model. Wiley.
  • McKinsey & Company
    • (2025). The State of AI: How Organizations Are Rewiring to Capture Value. McKinsey QuantumBlack.
    • (2025). How an AI-enabled software product development life cycle will fuel innovation. McKinsey Technology.
  • Palacio, J. (2026). Scrum in the Age of AI: The Synthesis to Come. Navegápolis.
  • Torres, T. (2021). Continuous Discovery Habits. Product Talk LLC.
  • Wolpers, S. (2026). AI4Agile Practitioners Report 2026. Age-of-Product / Scrum.org.

Take part in the poll

This article includes an active poll in the Scrum Manager community.

poll View the poll in the community