Case 01
Thomson Reuters
The Enterprise AI Backbone — Human-in-the-Loop at Scale
The Challenge
Scale AI across revenue-generating editorial, legal, and tax research products and internal operations (finance, HR, legal ops) without compromising the accuracy and trust customers and regulators expect from a global information leader.
The Strategy
Stood up a federated AI platform — operated by an internal center of excellence — that embeds agentic workflows into both customer-facing product pipelines and back-office operations. Designed human-in-the-loop verification gates at every high-stakes step, so subject-matter experts review, correct, and sign off on AI outputs before they reach customers or financial systems. Multi-model orchestration across Anthropic, OpenAI, and Google forces structured disagreement on ambiguous cases; governance and observability convert each human correction into a continuous improvement signal for prompts and models.
The Outcome
- Reusable AI capabilities deployed to 13,000+ employees across product and internal teams
- Faster cycle times on revenue-generating research and editorial workflows with expert-verified accuracy
- Internal segments (finance, legal ops, HR) accelerated through HITL safeguards — reducing review burden without removing accountability
- Secured substantial executive investment backed by measurable productivity and quality gains