Insurance Prior Auth Copilot | NTS Healthcare AI & ERP

Healthcare AI & ERP / Revenue cycle

Before many tests, procedures and medications, insurers require prior authorization: proof that the patient meets the insurer's coverage criteria. ClearAuth takes a patient's clinical documents and the insurer's policy, extracts the relevant facts, and matches each piece of evidence to each policy requirement. The result is an evidence matrix showing what is satisfied and what is missing, a Documentation Readiness Score, and a draft authorization packet that can be exported as a printable report or in a FHIR R4-style format.

It is built for the hospital clinical and revenue-cycle staff who prepare authorization requests. They pull together scattered notes, labs, imaging, treatment history and codes, and check them against insurer criteria that change often. Working from a dashboard, a case queue, a guided intake and a case workspace, a reviewer sees each extracted fact next to its exact source quote and page, confirms or rejects it, finds gaps before submission and generates a packet. A policy library lets teams upload, version and search insurer policies.

ClearAuth is citation-first. Rather than letting an AI model answer from memory, it first retrieves the active, date-appropriate policy sections and then allows citations only to what it retrieved. Clinical facts keep their full provenance from upload to final packet: exact text, page, document, extraction method, confidence, and whether a person has verified them. Packets are checked so that unsupported claims are rejected. The Readiness Score measures documentation completeness only and is explicitly not a prediction of approval. The workflow uses a single, predictable orchestrator with small tools rather than autonomous agents, so its behavior is easy to inspect.

A reproducible evaluation suite runs ten synthetic failure scenarios against the real retrieval, matching, citation and export code, measuring policy-retrieval recall, extraction precision and recall, citation validity, unsupported-claim rate and missing-information detection. It scores perfectly on that set, and the project is clear that these are regression results on a small synthetic dataset, not clinical validation. ClearAuth never approves or denies coverage, makes medical decisions or guarantees authorization; a human reviewer stays accountable for every submission. AWS infrastructure (API Gateway, ECS, Aurora, S3, OpenSearch, Bedrock, KMS) is scaffolded in Terraform, and the documentation includes a candid analysis of what production would still require.

The Insurance Prior Auth Copilot dashboard: active, draft, information-required and ready-for-review cases, average processing time, recently updated cases with their documentation readiness, and the readiness distribution, on synthetic demonstration data.

The Insurance Prior Auth Copilot authorization cases list: synthetic cases with status, payer and plan, requested service and code, urgency, assigned reviewer and documentation readiness, with search and filters. — The case list shows every authorization request with its payer and plan, the requested service and code, urgency, assigned reviewer and status. Each case carries a Documentation Readiness Score, which measures how complete the documentation is against the policy, not how likely the request is to be approved.

The Insurance Prior Auth Copilot policy library: synthetic payer policies with their versions and verification status, and a preview of one policy's details and documentation requirements. — The policy library stores each payer policy with its version, plan and effective dates, and reviewers can edit its details or reindex it for search. The preview shows the documentation requirements that the copilot retrieves and cites when it checks a case against this policy.

Start a free trial

Fill in the details below and we'll set up your free trial of Insurance Prior Auth Copilot on synthetic data and email you access.