AI operating model
What is AI operating model?
How AI work is owned, funded and governed across the company: who decides, who builds, who sets the standards. The common pattern is hub-and-spoke, a central team holding platforms and rules while business units run their own use cases. It's the difference between scattered pilots and a programme that compounds value.
Why it matters
The question an operating model answers sounds dull and turns out to be decisive: when a business unit wants to build something with AI, who do they ask, who pays, and who can say no. Get it wrong in either direction and progress stops. Push everything through a central team and the queue backs up while the people closest to the problem wait for permission. Scatter it entirely and every unit reinvents the same security review, buys overlapping tools and shares nothing. Most companies land in between, and the balance moves as they mature, with tighter central control early, when skills and standards are scarce, and more local freedom later.
In practice
Early on, a central group owns the platform, vets tools and trains the first teams, because letting 40 people each pick their own stack would be chaos. As those teams grow competent, the same group hands more freedom to the units and shifts to setting guardrails instead of approving every request. The structure is not fixed at launch. It loosens as trust and skill spread.
