Underwriting Journey

This page walks through the complete Silverarrow underwriting journey — from the moment a broker sends a submission email to the final exported risk analysis.

1. Receive

Silverarrow connects to your organisation’s mailbox (e.g., Microsoft Outlook via Graph API) and monitors for new submission emails. Thee system polls at short intervals and once a new submission email is identified:

  • The email metadata (sender, subject, date, body) is captured
  • The proposed risk structure — including, for example, layering — is identified from the submission email itself
  • All attachments are downloaded and stored securely; these attachments form the document basis for the case
  • A new case is created
  • A risk appetite check is performed against configured industries, treaty conditions, and underwriting guidelines

You do not need to manually forward or upload emails. The system handles ingestion automatically once your mailbox is connected.

Attachments

For a case to proceed through the full analysis pipeline, the following attachments are required, after passing the risk appetite check:

  • Submission email — Always provided as the entry point; captured automatically. Must contain the proposed risk structure (e.g. layering)
  • Primary policy — The current in-force policy or, where already available, the upcoming signed policy (the latter is preferred as it reflects the most recent agreed terms)
  • Master claims history / histories — Historical claims data covering the insured risk

The following documents are not required but enrich the analysis when available:

  • Submission — A MECE document about the client covering all relevant underwriting information, with the exception of the detailed claims history. Where provided, this is the most complete single source of client information
  • Sanctions questionnaire
  • Other specific questionnaires (e.g., product liability, environmental)
  • USA claims histories — Where US exposure is present

Missing required documents are automatically requested from the broker. Should the automatic request not yield the necessary documents, the underwriter can upload them manually from the submission detail view.

Document Classification

When documents are ingested, Silverarrow classifies them automatically using a two-tier approach:

  • Rule-based classification — File names, content patterns, and structural indicators are checked against known patterns. If the confidence score exceeds the threshold, the classification is applied immediately
  • AI fallback — If rule-based classification produces a low confidence score, the document is sent to an AI model for classification. The model reads the document content and assigns a type

If a document is classified incorrectly, the underwriter can manually reclassify it from the submission detail view. After reclassification, any dependent analysis will re-run using the corrected document type.

2. Extract / Query

The AI reads each document and extracts structured data fields relevant to underwriting analysis. Extraction follows a three-tier approach:

  • Text extraction — PDF text is extracted directly using text parsing
  • OCR fallback — For scanned documents or ambiguous fields, optical character recognition is applied with targeted AI re-extraction
  • Vision fallback — For fields still ambiguous after OCR, page images are sent to an AI vision model for multimodal re-extraction

In parallel, Silverarrow queries external registries and data sources via APIs to enrich the extracted data — including company verification and broader business intelligence lookups.

Every extracted field includes an evidence reference pointing back to its source document and location.

3. Structure

Silverarrow consolidates the extracted and queried data into a single, structured case view. All follow-up correspondence received after the initial submission is automatically associated with the same case, ensuring the underwriter always has the full picture in one place — with no data duplication across systems.

4. Analyse

With structured data in place, Silverarrow generates evidence-backed risk insights presented directly in the underwriter interface. Insights go beyond pure data structuring — they identify patterns, categorise findings into distinct topics, and surface relevant signals for the underwriter. These insights will eventually be tailored to the specific peculiarities of each market, including terminology, language, and scope.

5. Referral

When a case exceeds the underwriting authority of the handling underwriter, Silverarrow automatically initiates the referral process, routing the case to the appropriate authority level. This ensures that every submission is handled at the correct level without manual escalation.