The data function

In many organisations the data function has been treated as leftover work: something that feeds reports, unblocks a product team, or answers a board pack. It is rarely designed as a tier-one function, with a strategy, an architecture, and the same care given to finance, product, or operations.

When a function is neglected in that way, its systems grow by accident. Pipelines accumulate around the last urgent request. Reporting fragments by team. The architecture falls behind the business it is supposed to support, and the people closest to the data spend their time repairing last year's shortcuts.

I start from the opposite premise. The data function is an organisational capability. It has to support how the business operates, measures itself, and decides. That means designing it deliberately: architectural principles that hold, pipelines that can be reasoned about, governance people can actually follow, and a clear line from data work to the decisions it is for.

How data teams split work with AI is now part of that design, not a tooling afterthought.