Decentralize the building. Centralize the governance.

System connections

Executive intent
Operating model
Governance
Business outcome

Context

AI adoption was still fragmented: seven people had access, experimentation was ad hoc, and the organization needed a way to move from isolated use to durable enterprise capability.

The constraint was not simply access to better tools. In a regulated insurer, adoption had to account for security review, data classification, explainability, and responsible use while still giving business teams room to solve problems in their own domains.

Decision

I took ownership of the enterprise AI function and defined the strategy and operating model together. Business teams received real authority over build, buy, and hybrid decisions inside their domains.

The controls that needed consistency stayed central: security, data classification, explainability expectations, executive oversight, and responsible-AI governance. The operating model, not any single tool, was the capability being built.

Tradeoff

A single central AI team would have offered tighter control, but it would also have made that team the only build path and placed every business request into the same queue.

The choice favored speed and domain ownership over centralized execution. That only works when the guardrails are clear enough for teams to act without interpreting the rules differently each time.

Outcome

7 570 Users with governed enterprise AI access across eleven business units

Enterprise AI access scaled from 7 to 570 users across 11 business units.

The more important result was organizational: adoption and governance advanced together rather than being treated as opposing goals.

Evidence

The public claims in this case study are limited to approved wording and supported by the following published artifact.

Executive résumé