Most organisations are designed vertically. Customers experience them horizontally.
We create departments because they make accountability, expertise, budgets, and reporting manageable. There is nothing inherently wrong with that. The problem appears when we confuse the structure used to manage people with the way value is created.
A customer outcome usually crosses several functions, systems, and decision points. Each team may perform its part well while the overall experience remains slow, fragmented, or unclear. Local measures can improve while the end-to-end result does not. The gaps live in the hand-offs, and hand-offs rarely belong to anyone.
AI is making this more visible. An agent designed within one department may optimise that department’s task while creating additional work upstream or downstream. A faster assessment may overwhelm the approval step. Better demand generation may increase pressure on fulfilment. Automation can move a bottleneck rather than remove it.
That is why I have become more interested in value streams. A value stream gives us a way to see how work, information and decisions move towards an outcome. It does not eliminate functions. It creates a second lens through which we can manage the organisation.
I can imagine a transitional governance model built around Value Stream Councils. Senior people with genuine skin in the game would consider changes that cross process boundaries. Process owners would remain empowered to make contained improvements within agreed guardrails. The council would address changes with wider consequences, and a named Value Stream Owner would decide when consensus could not be reached.
Sending every improvement to a council would recreate the bureaucracy we are trying to remove. Allowing every process to optimise independently would retain the fragmentation. Authority needs to follow blast radius.
This is not only a governance question. It is an information question. To make good value-stream decisions, the organisation needs shared measures, visible dependencies, and an honest account of how work really happens. That takes business architecture, process ownership, and the candour to expose local problems without defending local territory.
Departments will remain important. But if we want AI to improve outcomes rather than accelerate isolated activity, we need to design and govern across them. Value does not stop at the edge of an organisational chart. Neither should our view of the work.