Decision owner
Underwriter
The named role remains accountable for the consequential decision.
Insurance
Designed for Underwriter
Turn unstructured broker submissions into a triaged, priced risk file by extracting exposures, normalizing loss runs, and testing appetite and guideline fit before an underwriter binds or declines.
Contract INS-02 · primary aggregate SubmissionFile · mode loading
Decision brief
This brief starts with the decision, evidence, and accountable owner. The guided workflow then shows how the recommendation was produced.
The evidence and control points that shape this recommendation.
Decision owner
Underwriter
The named role remains accountable for the consequential decision.
Governed evidence
5 record types
Structured records and cited text ground the recommendation.
Workflow
5 controlled stages
Each transition records its input, output, status, and audit context.
Human controls
1 approval gate
Approval gates prevent the copilot from making consequential decisions.
A broker emails a 900-location retail property submission with five years of loss runs; the workbench normalizes the schedule, geocodes locations into two wind-exposed counties, computes a burn cost that exceeds the appetite threshold, and the underwriter refers the account with the cited capacity clause rather than quoting it..
Turn unstructured broker submissions into a triaged, priced risk file by extracting exposures, normalizing loss runs, and testing appetite and guideline fit before an underwriter binds or declines.
Issued quote, referral to senior authority, or declination gives Underwriter the evidence needed to begin the final governed business outcome, with the decision trail retained for review.
Step 1 of 5
| Field | Value |
|---|
| Field | Value |
|---|
The output appears here immediately. More technical evidence remains available below.
The demonstration keeps deterministic controls, AI assistance, and approval separate.
Source systems enforce identity, workflow order, policy filters, and record updates.
Retrieval and generation assemble relevant evidence into a cited recommendation or draft.
Underwriter verifies material claims and approves the consequential action.
Run the stages in order before using the Copilot. Every stage explains what happens, why it matters, the evidence it uses, and the complete API request and response.
Automated stage
What happens: Submission intake service performs ingest broker submission package using Broker email with schedules and loss runs and produces SubmissionFile with attachment inventory. The transition is deterministic and does not require autonomous model approval.
Why it matters: SubmissionFile with attachment inventory gives Document extraction agent the evidence needed to begin Extract and normalize exposures and losses, with the decision trail retained for review.
Loading relevant sample records.
Ready to run.
POST /api/domain/workflow/ingest-broker-submission-package
Run this API to see its formatted outcome.
| Field | Value |
|---|
| Field | Value |
|---|
Automated stage
What happens: Document extraction agent performs extract and normalize exposures and losses using Spreadsheets and carrier loss-run PDFs and produces ExposureSchedule and LossRunRecord sets. The transition is deterministic and does not require autonomous model approval.
Why it matters: ExposureSchedule and LossRunRecord sets gives Triage agent the evidence needed to begin Score appetite fit and triage priority, with the decision trail retained for review.
Loading relevant sample records.
Ready to run.
POST /api/domain/workflow/extract-and-normalize-exposures-and-losses
Run this API to see its formatted outcome.
| Field | Value |
|---|
| Field | Value |
|---|
Automated stage
What happens: Triage agent performs score appetite fit and triage priority using Normalized exposures and guideline clauses and produces Triage score with referral reasons. The transition is deterministic and does not require autonomous model approval.
Why it matters: Triage score with referral reasons gives Underwriting analysis agent the evidence needed to begin Produce hazard summary and pricing indication, with the decision trail retained for review.
Loading relevant sample records.
Ready to run.
POST /api/domain/workflow/score-appetite-fit-and-triage-priority
Run this API to see its formatted outcome.
| Field | Value |
|---|
| Field | Value |
|---|
Automated stage
What happens: Underwriting analysis agent performs produce hazard summary and pricing indication using Burn cost, hazard data, capacity limits and produces RiskAssessment with cited rationale. The transition is deterministic and does not require autonomous model approval.
Why it matters: RiskAssessment with cited rationale gives Underwriter the evidence needed to begin Quote, refer, or decline, with the decision trail retained for review.
Loading relevant sample records.
Ready to run.
POST /api/domain/workflow/produce-hazard-summary-and-pricing-indication
Run this API to see its formatted outcome.
| Field | Value |
|---|
| Field | Value |
|---|
Human approval required
What happens: Underwriter performs quote, refer, or decline using Risk assessment and pricing indication and produces Issued quote, referral to senior authority, or declination. A named human approver must review the evidence before the transition completes.
Why it matters: Issued quote, referral to senior authority, or declination gives Underwriter the evidence needed to begin the final governed business outcome, with the decision trail retained for review.
Loading relevant sample records.
Ready to run.
POST /api/domain/workflow/quote-refer-or-decline
Run this API to see its formatted outcome.
| Field | Value |
|---|
| Field | Value |
|---|
Run a stage to see its explanation, evidence, full request, and full response.
Ask a question and receive an answer grounded in retrieved SubmissionFile evidence.
Select a starter question or enter your own.
Loaded live from /api/config, /api/domain/model, and
/api/domain/sample.
submissionFiles/tenantId + /submissionIdA broker emails a 900-location retail property submission with five years of loss runs; the workbench normalizes the schedule, geocodes locations into two wind-exposed counties, computes a burn cost that exceeds the appetite threshold, and the underwriter refers the account with the cited capacity clause rather than quoting it.
| Container | Role | Hierarchical partition key | Document types |
|---|---|---|---|
submissionFiles |
Primary | /tenantId + /submissionId |
SubmissionFile, ExposureSchedule, LossRunRecord, RiskAssessment |
underwritingGuidelineLibrary |
Reference | /tenantId + /guidelineId |
UnderwritingGuideline |
| Type | Container | Partition key | Identifier pattern | Embedding fields |
|---|---|---|---|---|
SubmissionFile |
submissionFiles |
/tenantId + /submissionId |
submission::{submissionId} |
None |
ExposureSchedule |
submissionFiles |
/tenantId + /submissionId |
exposure::{submissionId}::{scheduleId} |
None |
LossRunRecord |
submissionFiles |
/tenantId + /submissionId |
lossrun::{submissionId}::{priorClaimId} |
None |
UnderwritingGuideline |
underwritingGuidelineLibrary |
/tenantId + /guidelineId |
guideline::{guidelineId}::{clauseIndex} |
clauseVector |
RiskAssessment |
submissionFiles |
/tenantId + /submissionId |
assessment::{submissionId}::{assessmentVersion} |
hazardSummaryVector |
Load schedules, loss runs, and the assessment for one submission in one partition read.
submissionFiles
single-partition
SELECT * FROM c WHERE c.tenantId = @tenantId AND c.submissionId = @submissionId AND c.type IN ('SubmissionFile','ExposureSchedule','LossRunRecord','RiskAssessment') ORDER BY c.assessmentVersion DESC
Partition-key fan-out control-plane query: prioritize undecided submissions by effective date and triage score. It cannot carry the second partition key because an underwriter's queue spans every submission assigned to them, so it runs as a tenant-scoped scheduled sweep off the interactive path, bounded by submissions in triage whose target effective date falls inside 30 days, executed with MaxItemCount paging and continuation tokens.
submissionFiles
partition-key-fan-out
SELECT * FROM c WHERE c.tenantId = @tenantId AND c.type = 'SubmissionFile' AND c.decisionState = 'inTriage' AND c.targetEffectiveUtc < @horizonUtc ORDER BY c.triageScore DESC
Identify submissions with large incurred losses in the recent policy years.
submissionFiles
single-partition
SELECT * FROM c WHERE c.tenantId = @tenantId AND c.submissionId = @submissionId AND c.type = 'LossRunRecord' AND c.incurredAmount > @largeLossThreshold AND c.policyYear >= @lookbackYear ORDER BY c.incurredAmount DESC
Partition-key fan-out control-plane query: route schedules whose extraction confidence is too low for automated pricing. It cannot carry the second partition key because the extraction quality queue is a pipeline-wide control surface, so it runs as a tenant-scoped scheduled sweep off the interactive path, bounded by schedules below the confidence floor on submissions still in triage, executed with MaxItemCount paging and continuation tokens.
submissionFiles
partition-key-fan-out
SELECT * FROM c WHERE c.tenantId = @tenantId AND c.type = 'ExposureSchedule' AND c.extractionConfidence < @confidenceFloor ORDER BY c.extractionConfidence ASC