Health-Claims Automation at an APAC Health Insurer — 74% Auto-Adjudication, Median Settlement in 6 Hours
Claims Router + Medical Records Parser + DocuMage automating the inbound claims workflow at one of the region's largest private health insurers.
The challenge
The insurer — one of the largest private health insurers in its APAC market with approximately 4.8 million members and a claims volume of approximately 320,000 per month — was operating a claims-handling function whose customer-experience metrics had been trending downward. Average claim settlement was 9 calendar days for hospital claims, 5 days for ambulatory claims and 3 days for pharmacy claims. The insurer's customer-experience research consistently identified the claims journey as the lowest-rated touchpoint, with members particularly frustrated by the delay between hospital discharge and claim settlement.
The claims-handling team — approximately 480 staff across two operations centres — was running structurally behind on volume that was growing 12% year-over-year. Auto-adjudication rate on the existing rules-engine was approximately 22%, meaning the bulk of claims required manual review by claims handlers. Manual-review productivity was constrained by the document-handling burden: each hospital claim arrived with a mix of structured EDI claims data, scanned hospital invoices, scanned clinical-documentation extracts (where required for medical-necessity assessment) and patient-supplied supporting documents.
The constraints were operational. The local-jurisdiction data-protection requirements (a regional equivalent of HIPAA with stricter cross-border-flow restrictions) meant any AI processing of medical data had to happen inside the insurer's locally-hosted infrastructure. The insurer's existing claims-processing platform — a customised commercial deployment — could not be replaced. The claims-handler team's institutional knowledge had to be captured and embedded into the automation rather than simply replaced.
The approach
MindMap deployed Claims Router (Cr) as the adjudication-and-routing engine, Medical Records Parser (Mp) for the clinical-documentation analysis, DocuMage for the invoice and supporting-document extraction, and Anomaly Detector (Ad) for the fraud and abuse layer.
Phase one was the policy-and-benefit knowledge-base build. The insurer's policy and benefit library — including the network-provider contracts, the per-product coverage terms, the medical-necessity criteria and the case-management rules — was migrated to Claims Router's structured policy store. Each policy element was decomposed into a machine-readable form that the adjudication engine could evaluate, with the claims-handling team's institutional knowledge captured as the per-policy-element rationale.
Phase two was the document-intelligence build. DocuMage was trained on the insurer's claims-document corpus — approximately 18 million historical claims documents — to extract the structured fields the adjudication engine requires. Hospital invoices, with their procedure-line-item detail, are the most-volume-intensive document type and received the deepest model tuning. Medical Records Parser handles the clinical-documentation extracts where medical-necessity assessment is required.
Phase three was the adjudication-and-handler-workflow build. For each inbound claim, the platform performs the document extraction, applies the policy-and-benefit rules, performs the network-provider contract checks, performs the fraud screening, and produces an adjudication outcome. Claims that clearly meet the adjudication criteria are auto-settled. Claims that clearly do not meet the criteria are auto-denied with the specific deficiency cited. Claims that require human review are routed to the appropriate handler queue with the platform's structured assessment as a starting point.
The pre-built building blocks
Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.
Claims Router
Adjudication-and-routing engine with policy-and-benefit knowledge base
Medical Records Parser
Clinical-documentation analysis for medical-necessity assessment
DocuMage
Hospital-invoice and supporting-document extraction
Anomaly Detector
Fraud-and-abuse pattern detection with graph-level analysis
The architecture
The platform runs entirely on the insurer's local-jurisdiction private cloud, with full local data-residency and the local-jurisdiction medical-data-protection posture maintained throughout. No medical data leaves the insurer's locally-hosted infrastructure.
Claims Router's adjudication engine is rule-based with ML augmentation. The policy-and-benefit rules are deterministic — the insurer's coverage rules are contractual and require deterministic application. The ML augmentation comes in for the assessment of free-text components (medical-necessity narratives, complaint-letter contextualisation) and for the fraud-and-anomaly layer.
DocuMage and Medical Records Parser use a fine-tuned Llama 3.1 70B variant trained on the insurer's de-identified claims-and-clinical-document corpus. The fine-tuning corpus is approximately 1.4 million labelled documents across the insurer's claim-document type set; the resulting model handles the insurer's specific document templates and clinical-vocabulary patterns at materially higher accuracy than a generic medical-document model.
Anomaly Detector's fraud-and-abuse layer combines deterministic rules (the standard healthcare-fraud-detection patterns — billing-pattern anomalies, network-leakage patterns, prescription-fraud patterns) with a graph-network analysis that identifies coordinated patterns across providers, patients and prescriber networks.
Integration with the insurer's existing claims-processing platform is via the platform's standard inbound and outbound APIs. The new adjudication outcomes are written back to the existing platform for downstream processing (payment release, member notification, provider notification). The platform sits upstream of the existing platform and accelerates the work that feeds it.
The full claims-processing audit trail — every document extraction, every rule evaluation, every model decision, every handler action — is persisted with the local-jurisdiction-required retention.
The numbers behind the story
Auto-adjudication rate has risen from 22% to 74% across the claim portfolio. Median settlement time has dropped from 9 calendar days (hospital), 5 days (ambulatory) and 3 days (pharmacy) to a portfolio-average of 6 hours from claim receipt to settlement decision.
Customer NPS on the claims journey has risen from 6.1 to 9.4 on the insurer's measurement scale. The settlement-speed improvement is the dominant driver — members who had expected a multi-week claims process are seeing settlement decisions within hours.
Claims-handler productivity on the cases that still require manual review has improved approximately 3x. The platform's structured assessment, the document extraction and the recommended adjudication eliminate most of the document-review and policy-lookup work the handler previously performed. Handler capacity has been redirected to complex-claim handling, fraud investigation and member-experience case work.
Fraud detection has improved as a side effect. The Anomaly Detector layer has identified categories of fraud and abuse that the previous rules-based approach systematically missed — particularly the cross-provider coordinated patterns that benefit from graph-level analysis. The insurer's fraud-investigation team's referral acceptance rate is up.
Cost-per-claim has dropped meaningfully. The combined effect of the auto-adjudication, the handler-productivity improvement on the remaining cases, and the fraud-loss reduction has produced a multi-year-payback business case that the insurer's CFO has been able to track in detail.
“Nine days to settle a hospital claim was the largest single driver of our customer-experience pain. MindMap delivered six-hour median settlement in twenty-eight weeks, with seventy-four per cent auto-adjudication and a customer-NPS uplift on the claims journey that exceeded our planning. Our claims-handler team is doing genuinely investigative work on the complex cases.”— Chief Claims Officer· APAC Health Insurer
Why MindMap was chosen
The insurer had evaluated three claims-automation vendors. Two were US-headquartered and built on US-claims-format assumptions; the third was a regional vendor whose document-intelligence depth was inadequate for the insurer's actual document mix.
MindMap's accelerator-composition approach — bringing Claims Router, Medical Records Parser, DocuMage and Anomaly Detector together into a unified claims platform with full local data-residency — was the structural differentiator. We could demonstrate the platform pattern at another APAC health insurer with comparable scope.
Our embedded APAC healthcare-claims expertise on the delivery team (three claims-operations veterans from peer APAC insurers and a clinically-trained delivery member for the medical-necessity components) was the third factor. The insurer's CCO felt that the team understood the operational realities of APAC health claims, not just the AI technology.
Related deployments
Prior Auth Acceleration
Prior Auth Accelerator + DocGenie automated 70% of payer prior auth submissions end-to-end, reducing turnaround from 3 days to 4 hours.
Medical Records Processing
Medical Records Parser processing 14,000 patient documents per day across nine hospitals, lifting coding accuracy from 87% to 99.2%.
Clinical-Trial Document Workflow
DocuMage + Clinical Trial Matcher accelerated study-start document cycle 5x — letting trial sites enrol patients weeks sooner than the prior baseline.
Want an outcome like this?
Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.