ECO-AI architecture

Agentic orchestration for regulated claims operations.

FNOL intake, document intelligence, policy interpretation, TIP and litigation review, reserve sequencing, and human-in-the-loop execution — on one product surface.

Talk to ClaimGPT API console

Core principle

The system recommends, drafts, and sequences. Licensed humans authorize.

Primary stack

MCP orchestration, multimodal parsing, core-claims APIs, audit-grade traceability.

Target systems

Guidewire, Duck Creek, Sapiens, fraud vendors, correspondence services.

MCP tooling

  • orchestrate_claim_workspace — one-call FNOL + widget
  • query_policy_system — coverage lookup
  • extract_document_data — multimodal parsing
  • run_fraud_analytics — anomaly scoring
  • update_claim_status — gated write-backs
  • draft_communication — adjuster-reviewed drafts

Workflow stages

  • FNOL triage and routing
  • Investigation and contradiction review
  • Policy interpretation
  • Settlement preparation

Guardrails

  • Traceable citations on material outputs
  • No unilateral denials or binding liability
  • HIPAA/GLBA-aware data minimization
  • RBAC and prompt-injection-conscious tools

Event-driven integration

Portal uploads, claim webhooks, or queue events trigger parsing and policy/fraud lookups. ClaimGPT builds a structured file narrative, flags missing prerequisites, and sequences the next API calls for the adjuster.

Why this branch exists

The ECO-AI workspace delivers claim-level decision support with TIP, litigation, and fraud views. The enterprise layer shows how the same site scales to carrier-grade orchestration and straight-through processing.

See it in product

Live workspace with guardrails, tool maps, and integration architecture.

Developer setup

Operator outcome

Adjusters see both the recommendation and the orchestration path on one surface.

Contact support Claims API