Cohere Health deployed an AI system that automatically reviews and approves health insurance prior authorization requests — the formal approvals insurers require before patients can receive certain treatments. The system handles up to 85–90% of requests in real time, with the remainder routed to human clinical reviewers.

The platform uses machine learning (ML), natural language processing (NLP), and optical character recognition (OCR) to ingest authorization requests submitted via phone, fax, portal, API, or EHR, extract relevant clinical evidence from structured and unstructured documents, and compare the request against evidence-based clinical guidelines and the health plan's policies. When the clinical data supports approval, the AI issues an automated decision; complex or ambiguous cases are escalated to on-staff nurses or specialty physicians. Cohere processes over 12 million prior authorization requests annually, working with over 660,000 providers and health plans in all 50 states. Prior to AI-based systems, Cohere's Chief Medical Officer has publicly described the standard industry process as a purely manual review where a nurse or physician had to read through all submitted documents, a slow process with an estimated 10% error rate.