Most companies don't lack an AI strategy because nobody's thought about it. They lack one because the thinking never got installed as an actual operating structure — a hub with a mandate, a cadence, and the authority to say no.

A thin platform hub isn't a committee and it isn't a full center of excellence with its own P&L. It's a small, senior team — often four to eight people — sitting at the intersection of engineering, security, legal, and the business, with the explicit mandate to own provisioning, vendor relationships, and the shared infrastructure that every team building on AI actually needs.

The first 90 days matter disproportionately, because that's the window where the hub either earns real authority or gets treated as a formality everyone routes around. That means shipping something people actually feel in the first month — usually governed, fast provisioning that's a genuine upgrade over whatever ad hoc access existed before — not spending the whole quarter writing a charter nobody reads.

Adoption is near-universal. Governed operation is not. That gap is where risk and wasted money concentrate.

By day 90, a real platform hub should be able to name every production AI system in the company, who owns it, what it costs, and what happens if the underlying model gets deprecated next week. Most companies can't answer that today. That's the gap a hub with real mandate closes.