There is a familiar story in healthcare circles about Jos de Blok, a Dutch nurse who saw how management structures could get in the way of caring for patients. In 2006, after working as both a nurse and a manager, he got tired of the inefficiencies of the structure, helped start Buurtzorg with four nurses who could organize care around their patients' needs. 1
The organization grew to thousands of employees while keeping teams small. Each has at most twelve nurses and splits when it needs to grow. The nurses organize their work, including scheduling and recruitment, with shared technology, administration, and coaches for support. 2
What interests me is the arrangement around those teams. They had responsibility for the work, authority to organize it, and support they could draw on. That combination makes Buurtzorg useful to think about as AI expands what a small team can accomplish.
One study finds that AI reinforces what an organization does well and makes its existing problems worse. The gains depend on improving the way work is organized. 3 If teams already struggle to agree on priorities or combine their work, producing more can deepen those problems.
With agents, each person can research, prototype, and test more ideas. In a large team, people may then make changes faster than they can resolve conflicts between them. Keeping the group small reduces how many people must agree on the same decisions. As each person becomes more capable, the team can take on more without adding more people to coordinate. That makes a good case for small teams. But keeping teams small is not enough.
A small team needs the freedom to improve how the work gets done. If it is still expected to follow the same procedures and complete the same list of tasks, AI may simply help it do unnecessary work faster. John Seddon makes a similar point in Freedom from Command and Control: empowering people achieves little if the way work is organized stays the same. 4 If customers keep asking for help with the same form, AI can help the team answer faster. The team also needs authority to simplify and fix the form so customers can complete it without help.
Additionally, there is a need for freedom from team dependencies. DORA links better delivery to teams that can test and deliver product changes independently without waiting for several other teams to do their part. That takes more than permission: the creative process must allow one part to change without forcing changes elsewhere. 5 Shared tools and access to specialists help make that independence practical. But there is another essential ingredient for success.
As a small team takes on work once spread across several roles, the organization must trust its judgment across those roles. Contributors face a similar shift with agents: they may rely on work they cannot inspect in every detail. They need to judge where close review matters and where tests and feedback provide enough evidence to continue.
That makes early feedback essential. Research finds stronger benefits from AI, when teams work in small batches. Smaller changes let teams learn sooner and correct an assumption before more work depends on it. 6 But short cycles only help if the team can act on what it learns. Testing a prototype may reveal a problem that requires changing the design, the underlying systems, or how customer data is protected. The team needs authority to follow the problem across those boundaries.
A year and a half ago, I joined Massproductive, a small agency where we are trying to give small teams the freedom to make decisions and accomplish more with AI assistance. We are seeing the benefits of applying these principles, and learning from it.
Going back to story of Buurtzorg, keeping teams small meant letting them organize care, with support around them. For us, the same principle means moving decisions into the team as its responsibilities expand. Established organizations will have to give up some departmental control for that to happen. What organizational changes are you making to augment your team with AI?
SHRM (2024). In First Person: Jos de Blok.
Martela & Nandram (2025). Buurtzorg: Scaling up an organization with hundreds of self-managing teams but no middle managers.
DORA (2025). State of AI-assisted Software Development.
Seddon (2005). Freedom from Command and Control: Rethinking Management for Lean Service. Introduction, pp. xii–xvi.
DORA. Loosely Coupled Teams.
DORA. Working in Small Batches.
