How it works
AI scaling. Rule-based precision.
A multi-stage ingestion engine that runs deterministic rules before AI, validates everything against the Benefit Plan Standard, and binds every value to its source.
The Ingestion Pipeline
We don't just dump text. Our pipeline treats plan documents as data sources. We identify SBCs, EOCs, and Riders, extract their structure with an independent layout engine, and map them to a canonical vocabulary through deterministic rules. AI is used where it helps, never where it would have to be trusted blindly.
Traceability is built-in: every output field carries a provenance record pointing to the page, bounding box, and source text in the original PDF. Built for audit trails and dispute resolution.
Normalized Output
Forget parsing "Ded Waived" vs "No Ded". We normalize carrier quirks into a clean, typed schema, conformant to BPS v1.1:
- Accumulators: Deductibles & MOOPs split by network tier.
- Cost Shares: Copays, coinsurance, and deductible applicability.
- Limits: Structured visit counts and frequency rules.
- Pharmacy: Formulary tiers, networks, and coverage conditions.
Pharmacy & Formulary
Drug coverage, structured and cited.
The newest module in the engine, built on the Benefit Plan Standard's pharmacy module. Live today, not on a roadmap.
Formulary tiers, typed
Carrier drug classes map to canonical tier codes, with retail and mail-order networks modeled separately. Generic, preferred brand, non-preferred, and specialty tiers come out as structured data.
Split prices, split citations
Real SBCs put three prices in one table cell. We split them into labeled sub-tiers, and each price cites its own line in the source document. One cell, three prices, three distinct citations.
Coverage conditions
Prior authorization, step therapy, quantity limits, and indication-dependent coverage are captured as structured conditions. For GLP-1s and other high-cost drugs, that nuance is the answer.
Key Capabilities
Built for real health plan workflows.
Our platform focuses on features that actually matter to provider groups, digital health platforms, brokers, TPAs, and insurtech products.
Intelligent Ingestion
Send us PDFs, Word docs, or HTML. Our classifiers automatically detect the document type and apply the correct extraction strategy, combining AI document analysis with deterministic rules.
Deep Normalization
One schema across every carrier. Curated crosswalks handle the vocabulary quirks of the major national and regional carriers, so "Ded/Coins" and "deductible then coinsurance" land in the same typed field.
Enterprise Security
Built on Azure with private storage and environment isolation. For enterprise clients, your proprietary plan documents never mingle with public data.
Traceable Artifacts
Trust, but verify. We expose the raw, extracted, mapped, and validation artifacts so your compliance team can audit every data point end to end.
Cost-Share Modeling
We structure financial data (deductibles, OOP max, coinsurance) so you can actually run math on it. Perfect for quoting engines.
Portal, API, and MCP
The same cited answers through three surfaces: a Portal for ops teams, a REST API for production systems, and an MCP Server for AI agents. One engine underneath all three.
Standards alignment
Health Plan API serves documents conformant to the Benefit Plan Standard (v1.1, with the pharmacy module), an open standard with documented alignment to FHIR, including InsurancePlan and the CARIN Digital Insurance Card work. BPS handles what happens before FHIR: turning unstructured plan documents into normalized, provenance-carrying data that FHIR-based exchange can build on.
Roadmap
Pharmacy and formulary support is live. Next: behavioral health, dental & vision, and supplemental benefits, aligned with the Benefit Plan Standard roadmap.
Design partners help shape carrier priority, module scope, and quality guarantees. If you have a specific carrier or product line you need, we'd love to hear from you.