An end-to-end underwriting copilot experience that combines a web UI with an agentic backend to accelerate case review, surface key risk signals, and produce explainable AI decision support.
- Automated risk scoring and faster quote decisions
- Human review guidance and routing for efficiency and control
- Transparent recommendations with explainability to support review
- Better alignment to policy and compliance expectations
- Web experience optimized for underwriting workflows (case review + AI insights)
- API-first backend enabling copilot-style interactions
- Explainability-oriented outputs (how/why conclusions were produced)
- Case review experience with structured sections
- AI-assisted summaries and decision support
- Location/property intelligence and risk signals (where available)
This section shows two views:
- Logical architecture: a high-level view of how data sources feed an agent system to produce user-facing outcomes.
- Reference architecture: a more detailed view showing the UI workbench and agent roles.
- Open the deployed UI URL.
- Select a case and review the available panels.
- Trigger AI/copilot actions as exposed in the UI.
- Review outputs and validate against your underwriting process.
This repository intentionally avoids publishing environment-specific URLs and secrets.
- Deployed UI endpoint (Azure App Service)
- Deployed backend endpoint (Azure App Service)
- Appropriate access to any required AI provider (credentials configured via environment variables in the deployment environment)
agentic-underwriting-backend/– Backend API service (FastAPI)agentic-underwriting-ui/– Frontend web app (Next.js)
- Backend technical docs:
agentic-underwriting-backend/README.md - UI technical docs:
agentic-underwriting-ui/README.md
- Do not commit secrets (API keys, endpoints, tokens).
- Use App Service Configuration / Key Vault (recommended) for secret management.
- Avoid sharing pilot data outside approved channels.