Automating MLR review for Comirnaty's global campaign

A global campaign meant the same claims moving through MLR review in many markets at once. Review ran in multiple rounds, was tracked by email, and depended on which reviewer picked up the work. The same claim could be validated differently from one region to the next, prior approvals were not reused, and there was no traceable record of why a decision was made.

• One global campaign, many market teams
• Multi-round, email-tracked review
• No decision traceability or reuse of prior approvals

I led the platform from pilot to production across markets. Content is ingested, its claims are extracted, and each claim is checked against an approved claims library using retrieval-augmented generation. A confidence score then sets the review route and service level, from fast-track to senior review for novel claims, and Workfront automation creates the review task, assigns the reviewer and tracks each decision to approval in place of email.

The design choices were deliberate: assist reviewers rather than replace them, show claim-level evidence, escalate whenever the model is uncertain, and keep an immutable decision history. At launch every decision still went to a human reviewer, which built the trust needed for adoption. I mapped responsibilities across content, AI automation, medical, legal, regulatory, and brand teams for every stage from drafting to publication.