Healthi
India
Healthcare Insurance
Amazon Bedrock, Amazon S3, Amazon SageMaker, Amazon DynamoDB, AWS Lambda, AWS IAM, Vector Database (RAG policy retrieval)
Healthi
India
Healthcare Insurance
Amazon Bedrock, Amazon S3, Amazon SageMaker, Amazon DynamoDB, AWS Lambda, AWS IAM, Vector Database (RAG policy retrieval)
Healthi, a healthcare technology company in India, wanted to modernize how small health insurance claims are approved at the point of care. Traditional pre-authorization created friction for routine, low-value transactions like consultations and prescriptions. Avahi built an agentdriven, real-time claims analytics system that digitizes claim documents, validates member and service details, retrieves policy clauses, and adjudicates the claim with an explanation and confidence score. The result was a faster, smoother member experience and a scalable path to on-the-spot insurance verification.
Healthi is a healthcare technology company in India focused on transforming medical documentation and claims processing through automation and Al-driven solutions. It supports digitization, interoperability, and workflow optimization for healthcare providers, insurers, and patients, improving operational efficiency across the healthcare ecosystem.
For common, low-cost claims, Healthi followed the industry norm of requiring pre-authorization before appointments or prescription orders. While workable for high-cost procedures, this approach produced a poor experience for routine transactions that can happen multiple times per year, adding delays and unnecessary steps for both members and providers.
Healthi needed a solution that could adjudicate claims in real time at the point of billing, so customers could immediately see whether a claim was covered and how much insurance would pay. To do this safely and accurately, the system had to digitize documents, verify the person and service details, confirm policy coverage, validate supporting documentation, and classify claims using ICD coding standards.
AWS provided the building blocks to ingest and securely store multimodal claim data (images and PDFs), orchestrate agent workflows with serverless services, and apply foundation models for policy understanding and claims reasoning. With managed services for storage, identity and access control, and scalable compute, Healthi could implement real-time adjudication with strong governance and a clear path to production scalability.
Healthi chose Avahi for deep experience designing agentic Al workflows that combine retrieval, document understanding, and rulebased validation. Avahi delivered an end-toend architecture that connected authentication, policy retrieval, claims digitization, coding classification, and adjudication into a single orchestrated flow. The solution also prioritized auditability by returning clear explanations and confidence signals for each decision, supporting trust and operational adoption.
Avahi developed an agent-driven workflow capable of adjudicating claims on the spot. The workflow begins with an Intent and Authentication Agent that verifies the user type and access permissions, enforcing role-based access via AWS IAM and controlling access to member policy data stored in DynamoDB.
Next, a Policy Retrieval Agent uses retrieval augmented generation to search the insurance policy corpus and extract relevant coverage clauses. Policy documents and claim artifacts are stored in Amazon S3, while a vector database supports semantic retrieval for faster and more accurate policy matching.
For intake, the Claims Intake and Digitization Agent processes multimodal claim inputs, including images and PDFs, performs OCR and structure extraction, and classifies the claim using ICD and CPT standards. A Real-Time Adjudication Agent then cross-validates the extracted claim details against policy rules and supporting documentation.
To ensure consistent outcomes, Avahi implemented an LLM-as-a-judge pattern that produces clear decisions, Approved, Partially Approved, or Declined, along with an explanation and confidence score. AWS Lambda orchestrates the end-to-end agent workflow, and a Streamlit dashboard provides a test and demonstration layer. Amazon Bedrock supports the LLM-driven reasoning, and Amazon SageMaker is used for model deployment and inference where needed.
Healthi can now provide real-time claim decisions at the point of billing, removing unnecessary pre-authorization steps for routine healthcare transactions. The workflow improves member experience while giving insurers and providers faster clarity on coverage, decision rationale, and confidence signals that support operational trust.
Let’s explore your high-impact AI opportunities together in a complimentary session
Let’s explore your high-impact AI opportunities together in a complimentary session
Let’s explore your high-impact AI opportunities together in a complimentary half-day session
