Prudential plc

Prudential plc, one of Asia and Africa's largest life and health insurers with more than 19 million life customers, faced mounting pressure from growing volumes and velocity of health insurance claims requiring manual review. Each claim submission typically involves multiple document types — diagnostic reports, prescriptions, and invoices — all requiring human data entry and interpretation before a decision can be reached. At scale, manual processing created compounding risks: higher error rates from transcription mistakes, inconsistent coding of medical information, and slower turnaround times for policyholders at exactly the moment they most need responsive service. The operational burden threatened both unit economics and customer trust.

Prudential partnered with Google Cloud to deploy MedLM — Google's family of foundation models specifically fine-tuned for healthcare industry use cases — against the document-heavy claims intake process. MedLM was applied to extract and accurately code relevant clinical and billing information from submitted documents, structuring unstructured medical data to support downstream adjudication. The deployment began as a controlled 3–4 month pilot in Singapore and Malaysia, running MedLM's analysis in parallel with existing approval workflows so outputs could be directly benchmarked against established human decisions. Prudential maintained a "human in the loop" at critical decision stages throughout, satisfying both regulatory expectations and internal risk governance requirements. Google Cloud's Karan Bajwa noted the approach was designed to empower Prudential's workforce rather than replace it.

Proof-of-concept testing demonstrated that MedLM doubled the automation rate of claim reviews and assessments compared to the prior manual baseline — a 2x improvement in straight-through processing. Accuracy of claims decisions also improved, reducing the error risk associated with manual data entry across high-volume document types. Qualitative outcomes included:

• Faster turnaround times for policyholders during claims — a high-stakes moment of truth for customer trust
• Higher throughput capacity, enabling Prudential to handle increased claim volume without proportional headcount growth
• Improved consistency in medical information coding across Singapore and Malaysia operations

The pilot's side-by-side comparison methodology gave Prudential a rigorous baseline for evaluating where MedLM delivered the greatest productivity lift.

• Healthcare-domain foundation models outperform general-purpose LLMs for clinical document extraction; vertical-specific fine-tuning is a meaningful differentiator in regulated insurance workflows.
• A time-boxed parallel-run pilot (running AI outputs alongside existing decisions) creates a defensible evidence base before committing to full rollout.
• "Human in the loop" is not just a risk control — in regulated markets it is a prerequisite for deploying generative AI in claims adjudication.
• First-mover advantage in AI adoption can translate to operational efficiency gains that compound as claim volumes grow.
• Framing AI as a decision-support tool for assessors, rather than an autonomous replacement, reduces internal adoption friction and regulatory scrutiny.