The next generation of operating models will not have an AI strategy. They will have an operating model in which AI is a standing assumption — in how products are built, in how decisions are made, and in how the organisation reorganises around the work it can now do.
The argument
An AI-native operating model is recognisable by three structural shifts: product teams that own a model as a first-class artefact, a data function that has been collapsed back into the businesses it serves, and a governance layer that treats model risk the way the bank treats credit risk. Without all three, the organisation is running AI projects, not operating on AI.
What we see in the field
The companies moving fastest are not the ones with the largest AI teams. They are the ones where the head of a business line can sketch, on a napkin, the three model-mediated decisions that now run through their P&L every day — and name the person accountable for each.
What it changes
For boards, the implication is uncomfortable: the AI-native operating model cannot be delegated to a CIO or a chief AI officer. It is a structural choice about how the business is organised, and the cost of getting it wrong now compounds quarterly.
Where to start
Pick one P&L. Map the model-mediated decisions inside it today, the ones that should be model-mediated in twelve months, and the human authority for each. If the map can't be drawn, the operating model is not yet AI-native.
