An operating model is how work, money, and decisions flow. If you keep the silos and add squads, you have renamed the org chart. The customer will not notice the rename.
Value streams, not ceremonial squads
A value stream is the end-to-end path from a customer need to a customer result — including the handoffs, queues, and approval lakes that currently make it slow. Organising around that path means a standing team (or a small set of teams) can move a slice of value without raising a project to cross four departments.
Silos optimise local efficiency: a design queue, a build queue, a test queue, an operations queue. Each queue can look busy and healthy while the customer waits. Utilisation is a local trophy. Flow is the system prize. If your model still funds and manages the queues separately, the squads will spend their lives in alignment meetings.
You do not need to redraw the entire company on a Friday. Identify one stream, name an owner who can see the whole path, and give them enough of the skills inside the boundary that 'waiting on another department' becomes the exception. The rest of the organisation can remain ugly for a while. Ugly and honest beats pretty and fictional.
Product funding versus project packaging
Project packaging is: assemble strangers, deliver a scope, disband, hand a system to people who were not in the room. It fits unique, finite change. It is a poor default for a digital service that must evolve weekly. Product-shaped teams persist, own outcomes, and take work from a backlog ordered by value.
Hybrid organisations will have both. A plant move is a project. The customer portal is a product. The operating model should say which work is which, and stop forcing product teams through a project stage-gate designed for capital construction. That gate is how you pay an agility tax and still miss the season.
Decision rights sit in the model, not in a RACI nobody opens. Who can stop a release? Who can change the backlog order? Who can spend from the value-stream envelope? If those answers are 'a committee that meets monthly,' you have designed delay. If they are 'whoever shouted,' you have designed chaos. Write them at the seam.
Governance that matches the cadence
Predictive work wants baselines, change control, and stage gates that match irreversible cost. Adaptive work wants lightweight guardrails: architecture advice, security as a pipeline, a definition of done that includes compliance evidence, and a review that can redirect next increment. One governance flavour for both is how one of them suffocates.
Shared platforms need a model too. If every value stream rebuilds identity, payments, and data, you have funded a museum of slightly different wheels. A platform team with a service-level promise and an internal customer is an operating-model choice. A central team that is only a ticket queue is the old silo with a new Jira project.
Do not confuse a re-org with a model. Moving boxes on a slide does not change how money is requested or how success is measured. If those stay project-shaped, the boxes will drift back. We will take funding next, because money is how models become real or remain posters.
Summary & next steps
Organise around the customer path where you can, keep product teams persistent where the work is ongoing, and match governance to the type of work. A re-org without new decision rights is interior design.
Next: agile funding — the annual budget habit that quietly cancels the operating model you just drew.