The Operating Model: Pods, Not Projects
The quiet engine of every transformation McKinsey studied is the same: small cross-functional teams that own an outcome end to end. Change the org chart, or the technology won't matter.
Change the technology and leave the org chart alone, and the org chart wins. Every time. The quiet engine of these transformations is a different shape of team.
This is capability three, and the one people skip because it's the hardest — it asks you to change how teams are built, not just what tools they use. Across the transformations McKinsey studied, the same structure kept showing up: small, stable, cross-functional teams that own an outcome from end to end. The book calls them pods. A pod has everyone it needs to ship — product, engineering, design, data, the relevant business expertise — sitting together and accountable for a result, not a task. Contrast that with the model most companies still run:
Project — what fails
Specialists in separate departments hand work across walls — business → engineering → QA — then disband at launch. Nobody owns the outcome, because it arrives after everyone has scattered.
Pod — what good looks like
One persistent, whole team — product, eng, design, data, business — owns the metric, lives with what it ships, and stays to improve it.
Why ownership changes behaviour
When a team knows it will still be here in six months, living with whatever it ships, it makes different decisions. It stops optimising for “passed the hand-off” and starts optimising for “actually worked.” The shortcuts that look free in a project — the corner cut because it becomes someone else's problem — stop being free when there is no someone else. That's the deep reason pods outperform: end-to-end ownership removes the seams where accountability used to leak away, and the seams were where most of the value was being lost.
The agentic era is now bending this shape further. The leading edge is the agentic organisation — pods where humans and AI agents work inside the same workflow, the people setting direction and owning judgement while agents carry more of the execution. The unit doesn't change; the pod is still the answer. What changes is that a pod increasingly includes capability that isn't human, which makes the “one team owns the outcome” discipline matter more, not less — someone still has to own what the agents do.
Where it goes wrong
Renaming the project teams “pods” and changing nothing else. If the team still hands work to a separate department to finish, still disbands at launch, still answers to five managers with five priorities, it's a project wearing a new word. Half-measures here are why so many “agile transformations” produced new vocabulary and the same results.
Try this
Take one piece of work moving through your organisation and trace its hand-offs — count every time it crosses from one team or department to another before it reaches a user. Each crossing is a wall where ownership thins and value leaks. A pod is simply the bet that you'll get more by removing those walls than by managing the traffic across them.
Grounded in Lamarre, Smaje & Zemmel, Rewired (McKinsey), and McKinsey's writing on the agentic organisation.