A family business owner recently told me proudly: “We’re building our own AI operating model.”

A family business owner recently told me proudly: “We’re building our own AI operating model.”

He meant a set of rules for how AI works in the company. Who decides on use cases. Who can access which data. Which tool is approved. Who teaches people how to use it.

Three months of work.
An 80-page concept.
Not a single process worked differently than before.

The real problem was not the effort. It was the assumption behind it:

that a mid-sized company needs the same thing as a large corporation — a bespoke AI operating model built from scratch.

Most people see only two routes: build everything themselves. Or commission an expensive custom solution from a large consultancy.

There is a third.

Rather than reinvent everything, take three or four proven building blocks and adapt them:

– a small group that prioritises use cases
– clear data permissions
– an approved tool stack (e.g. Microsoft Copilot + a connected knowledge database)
– a training format that sticks

That is exactly what we later did at this company: adopted 80% proven architecture and adapted 20%.

Four weeks later, use-case approvals were settled, data access was defined — and the first AI programme went into production.

Not 100% built in-house. Not 100% a corporate blueprint.

Adapted beats bespoke. Almost always.

#MidMarket #Transformation #AI