The standard objection to self-organizing teams is that they are a luxury of software startups, unworkable where the stakes are public money, legal exposure, and scale. It is a testable claim. Two public-sector cases test it directly, and neither supports the objection.
A digital agency inside Whitehall
The United Kingdom's Government Digital Service is now studied as a paradigm case of a public digital agency. Rainer Kattel and Ville Takala (2023) document how GDS brought service design, agile practice, and a relentless focus on user need into central government, working in multidisciplinary teams rather than the traditional contract-and-specify model. The output most people have touched is GOV.UK. The output that mattered more was internal capability: the GDS Academy trained more than ten thousand civil servants in digital skills, which is how a way of working outlives the team that introduced it.
The interesting part for anyone worried that autonomy means chaos is how GDS paired freedom with accountability. Teams had real latitude, but services had to pass assessments at alpha, beta, and live stages against a published service standard, and GDS could stop a department from launching or spending further until those criteria were met (Kattel and Takala, 2023). This is aligned autonomy in a bureaucracy: teams decide the how, a central standard holds the why. It is the exact combination that the Spotify chart, copied so widely, left out. The standard was not a cage. It was the shared definition of done that let dozens of teams move independently without producing dozens of incompatible services.
The rescue of HealthCare.gov
The American case is more dramatic because it began as a public failure. The 2013 launch of HealthCare.gov collapsed under load, a visible embarrassment for an entire administration. What turned it around, by the government's own account, was not a bigger contract but a small set of interdisciplinary teams working in short iterations. The U.S. Digital Service (2016) reported concrete numbers to Congress. A rebuilt Scalable Login System reached 99.99 percent uptime and was described as far more stable than what it replaced. A simpler enrollment application, Marketplace Lite, raised the conversion rate to around 85 percent from roughly 55 percent in the prior system. Enrollments through the site rose from 5.3 million in March 2014 to 9.6 million in January 2016 as it stabilized. Small autonomous teams, shipping iteratively, inside a federal program with millions of users and no tolerance for failure.
Autonomy did not mean an absence of discipline
It would be a misreading to conclude that these teams succeeded because rules were removed. They succeeded because a specific discipline replaced the old one. The scaling-agile literature is blunt that transformations stall on coordination and management support far more than on team motivation (Dikert, Paasivaara and Lassenius, 2016). GDS answered that with a hard service standard and the authority to enforce it. USDS answered it with clear performance targets, uptime and conversion, that told everyone what good looked like. In both cases the center did not command the work. It defined the outcome and then trusted the teams to find the route, which is the difference between autonomy and abandonment.
That distinction is what most private-sector transformations miss. They copy the autonomy and forget the standard, or they keep the old command reflex and call the result agile. The public cases show a cleaner design: a small, well-supported multidisciplinary team, a published bar for what counts as done, and a center willing to hold that bar rather than micromanage the path to it.
The practical read
If autonomous teams can rescue a national health exchange and reshape how a government builds services, the burden of proof shifts to the leader who insists their own organization is too regulated, too large, or too important for the same approach. The public-sector evidence suggests the real prerequisites are not deregulation and slack. They are a clear standard of done, honest outcome metrics, and enough institutional backbone to hold both while the teams do the work (Kattel and Takala, 2023; U.S. Digital Service, 2016). Government cleared that bar. Most companies can too.
Sources
- Kattel and Takala (2023). The Case of the UK's Government Digital Service. Documents multidisciplinary agile teams in central government, paired with an enforced service standard and internal capability building. https://dl.acm.org/doi/10.1145/3630024
- U.S. Digital Service (2016). Report to Congress: Stabilizing and Improving HealthCare.gov. Government account of small interdisciplinary teams delivering 99.99 percent login uptime and an 85 percent enrollment conversion rate. https://www.usds.gov/report-to-congress/2016/healthcare-dot-gov/
- Dikert, Paasivaara and Lassenius (2016). Challenges and success factors for large-scale agile transformations. Finds transformations stall on coordination and management support rather than team motivation. https://doi.org/10.1016/j.jss.2016.06.013