AI Underwriting Intake Agent

How Cassidy automates Underwriting Intake using AI

Step 1: Trigger on new broker submission

The Workflow activates when a submission arrives—whether from an email inbox, broker portal, or SFTP drop—and consolidates related emails and attachments into a single submission package.

Step 2: Classify documents and check completeness

Cassidy auto-classifies document types (ACORDs, loss runs, SOVs, supplemental forms) against your requirements by line of business and jurisdiction, then flags missing items and drafts outstanding-requirements emails to brokers.

Step 3: Extract data to your carrier schema

Key fields—insured details, exposures, limits, class codes, COPE data, prior losses—are extracted into your defined schema with confidence scores and visual grounding back to source pages for human review when needed.

Step 4: Enrich and resolve entities

Cassidy augments submission data with third-party sources (geocoding, hazard indexes, firmographics, sanctions checks) and resolves entities across the email, ACORDs, and external data into a unified view.

Step 5: Run clearance and appetite checks

The Workflow performs broker-of-record validation, duplicate detection, and screens against your appetite and authority guidelines—auto-declining out-of-appetite risks and filtering to the right product or segment.

Step 6: Prioritize, route, and create referrals

Submissions are scored by fit, premium potential, and complexity, then routed to the appropriate underwriter queue. For exceptions outside authority, Cassidy auto-creates referrals in your PAS or workbench with full context, source docs, and audit trail attached.

Step 7: Pre-populate downstream systems

Extracted data flows directly into Guidewire, Duck Creek, or your rating engine—creating submission records, attaching documents, and enabling straight-through processing for low-complexity risks.