The agile promise was a team that governs itself: cross-functional, empowered, deciding how to do the work and, within limits, what to do next. Then came the product owner, the one throat to choke, the single voice for the backlog, the person who prioritizes and accepts and represents the business. In a healthy team the role clears obstacles and holds the "why." In an unhealthy one it becomes the place all the old authority quietly reassembled, a middle manager with a scrum certificate. The title changed. The reporting line to a single decider did not.
This is not an argument against the role. It is an observation about a pattern that organizational research predicts.
The boss who came back
Flat structures have a habit of regrowing their hierarchy. Foss and Klein (2022), studying supposedly bossless firms, found that many that claim to be flat in fact retain real bosses and hierarchies, just informal or renamed; formal hierarchy is replaced by informal hierarchy rather than eliminated. The product owner is a natural host for that regrowth. Someone has to say no to stakeholders, sequence the work, and accept what is done, and if the team does not hold those decisions collectively, they concentrate in the one person whose job description is closest to command.
The macro data suggests why the pull is so strong. Deloitte's 2025 Global Human Capital Trends, drawing on nearly 10,000 leaders across 93 countries, reports that managers spend close to 40 percent of their time on firefighting and administrative work and only about 13 percent developing their people, even as US postings for middle-management roles fell sharply. Organizations are trying to remove a layer of management and are discovering the coordination work that layer did still has to happen somewhere. In agile teams, it often lands on the product owner, who becomes the middle manager the org chart claims to have deleted.
The mechanism is subtle because it usually looks like help. The product owner shields the team from stakeholders, so the team stops talking to stakeholders. He owns the backlog, so the team stops deciding what matters most. Each act of protection removes a decision from the group and lodges it with an individual, and after a few months the team has the autonomy to choose its tools and the product owner holds the authority that counts. Nobody set out to build a hierarchy. One accreted, delegated decision by delegated decision, which is exactly the informal regrowth the research describes.
When authority should be shared
The alternative is not to abolish the role but to keep the decisions plural. The meta-analytic evidence favors distributing leadership rather than concentrating it. Wang, Waldman and Zhang (2014), pooling 42 samples, found an overall positive relationship between shared leadership and team effectiveness of rho = .34, and that the effect was stronger when the work was more complex, precisely the condition of software teams. D'Innocenzo, Mathieu and Kukenberger (2016), across 50 effect sizes and 3,198 teams, likewise found shared leadership positively related to team performance, with the effect strongest in real organizational settings. When a product owner monopolizes direction, a team forfeits the very structure the research associates with performance on complex work.
Semco pushed the logic to its limit by inverting the direction of accountability. As Semler has told it, everyone at the company evaluated their boss anonymously twice a year, and two low scores in a row meant losing the role, himself included. The team, not the boss's boss, could veto a manager. Managers who yelled learned to ask; those who could not adapt returned to strong technical work at the same pay. The lesson is not that product owners should be voted out. It is that authority over a team is legitimate only when the team has some hold over it. A product owner the team cannot push back on is not a servant leader. He is the old boss with a new backlog.
The diagnostic is simple. Ask where the team's real decisions live. If prioritization, acceptance, and direction are genuinely negotiated by the people doing the work, the product owner is doing the job the role was designed for. If they flow to and from one person who answers upward and not to the team, agile has not removed the middle manager. It has renamed him, seated him inside the circle, and called the circle flat.
Sources
- Foss & Klein (2022). Who's the boss? The persistence of entrepreneurial hierarchy in flat organizations. Firms that claim to be flat often retain real bosses; formal hierarchy is replaced by informal hierarchy, not eliminated. https://link.springer.com/article/10.1007/s41469-022-00132-8
- Deloitte (2025). 2025 Global Human Capital Trends (~10,000 leaders, 93 countries). Managers spend ~40% of time firefighting and only ~13% developing people, while US middle-management postings fell sharply. https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html
- Wang, Waldman & Zhang (2014). A meta-analysis of shared leadership and team effectiveness (42 samples). Shared leadership relates to effectiveness at rho = .34, more strongly when work is complex. https://pubmed.ncbi.nlm.nih.gov/24188392/
- D'Innocenzo, Mathieu & Kukenberger (2016). A Meta-Analysis of Shared Leadership-Team Performance Relations (50 effect sizes, 3,198 teams). Shared leadership is positively related to team performance, strongest in field settings. https://journals.sagepub.com/doi/10.1177/0149206314525205