The situation
The carrier had committed publicly to AI investment, but the strategy lived entirely in slides. There was no inventory of in-flight work, no governance process, no shared view of what data could be used for what purpose, and no clear path from pilots to production. The CTO and Chief Strategy Officer had a 90-day window before a board update.
What we did
We ran a 10-week engagement with three parallel workstreams:
- Current-state assessment. Capability mapping across business units, talent and tooling inventory, data and governance audit, and identification of in-flight pilots that did and did not justify continuation.
- Target architecture. A reference architecture covering platform, data, model, and governance layers — built so engineering could implement it and so risk could review it.
- Operating model. Two governance committees (technical review and risk review), a clear intake process, and explicit kill criteria so dead pilots could die without political cost.
Alongside the strategy work, we sequenced three use cases for a 60-day production push: claims triage assistance, underwriting research support, and an internal knowledge assistant for agents.
Outcome
The board approved the strategy and the funding to execute it. All three priority use cases shipped on schedule, with eval harnesses in place for ongoing review. The governance committees became the standing forum for new AI work — meaning the next request didn't need an outside firm to scope it.