A sound strategy and mature capabilities still fail if no one owns delivery and people do not actually use what is built. The operating model decides who does what, and adoption determines whether value is realised. Many AI programmes underdeliver not because the technology was weak but because ownership was unclear, skills were missing, or the workforce never changed how it worked.

Organisations choose among structures. A centre of excellence (CoE) centralises AI talent, standards, platforms, and governance in one team that serves the business, which is strong for consistency and reuse but can become a bottleneck. A federated (or hub-and-spoke) model embeds AI capability in business units with a central group setting standards and shared platforms, trading some consistency for speed and domain closeness. Many enterprises land on hub-and-spoke as a pragmatic middle path. There is no universally correct structure; the choice depends on maturity, scale, and how domain-specific the use cases are.

Beyond structure, adoption is its own discipline: change management (bringing people along, redesigning workflows, addressing fear), upskilling (building AI literacy broadly and deeper skills where needed), and measurement (tracking not just deployment but actual usage and realised value). A tool that ships but is not used delivers nothing, so adoption and value metrics, not model accuracy alone, are the real scorecard.

A manufacturer stands up a central CoE that builds excellent tools, but business units ignore them because they do not fit local workflows. Leadership shifts to a hub-and-spoke model: the central hub keeps platforms, standards, and governance, while embedded practitioners in each unit adapt solutions to local needs and drive adoption.

They pair this with change management (redesigning workflows and communicating the why), targeted upskilling, and a shift in metrics from "tools deployed" to "active usage and hours saved." Adoption climbs because capability sits close to the work while consistency is preserved centrally. The lesson: structure and adoption practices must fit the organisation, and value is measured by use, not by delivery.

The operating model operationalises everything else: it turns strategy into owned initiatives, is a dimension of maturity, enforces governance day to day, and decides who builds and monitors agents in the agentic landscape. Adoption is where value in the strategy is finally realised or lost. Without a fit-for-purpose operating model and deliberate adoption, even a strong, well-governed, mature strategy stalls at the last mile.