AI AgentUnderwriting Assistant
Copilot that summarizes risk, checks policy and drafts decisions
The challenge
Underwriters spend more time rekeying schedules and loss runs than assessing risk, so submissions sit in a queue and brokers place the business elsewhere. Appetite is applied inconsistently across the team, referrals arrive without a stated reason, and the basis for a decision is hard to reconstruct.
The outcome
Azure AI Document Intelligence turns schedules, surveys and loss runs into structured exposure data, Azure AI Search tests it against appetite and policy wordings, and Azure OpenAI drafts the risk summary and pricing rationale. A licensed underwriter still prices, agrees terms and binds every risk.
01 — Architecture
End-to-end architecture
A submission arrives from a broker, is read by Document Intelligence into structured exposure and loss data, and enriched with prior claims history. The Foundry agent tests the risk against appetite, limits and exclusions and drafts the terms — but a licensed underwriter prices, agrees and binds every risk.
- Broker: Broker / Applicant
- Submission: Azure Logic Apps, Azure Blob Storage
- Extraction: AI Document Intelligence, Azure Functions
- AI & assessment: Microsoft Foundry Agent, Azure OpenAI models, Azure AI Search, Azure AI Content Safety
- Data & records: Azure SQL, Azure Cosmos DB, Azure Key Vault
- Decision & insight: Underwriter decision, Communication Services, Power BI / Fabric
02 — Workflow
Process & decision workflow
How a submission becomes a decision-ready risk — extract, enrich, assess and draft, then branch. Risks inside appetite and inside authority get a drafted quote ready for release, while anything outside goes to an underwriter with the risk summary and the referral reason.
- Receive: Submission arrives by email, portal or broker feed
- Extract: Schedules, surveys and loss runs turned into data
- Enrich: Exposure, perils and prior claims history attached
- Assess: Risk tested against appetite, limits and exclusions
- Draft: Risk summary, terms and pricing rationale prepared
- Refer: Referral reasons and evidence packaged for review
- Path 1 · in appetite and authority — Draft quote for release: Terms and pricing prepared for issue
- Path 2 · outside appetite or limit — Underwriter decides: Risk summary, evidence and referral reason
03 — Components
Key Microsoft components
Underwriting is judgement supported by evidence — extraction, enrichment, appetite testing and audit all stay on the Microsoft stack.
Azure Logic AppsSubmission intake, triage, chasing and referral routing.
Azure AI Document IntelligenceReads schedules, surveys, valuations and loss runs.
Azure FunctionsNormalisation, exposure enrichment and rating inputs.
Microsoft Foundry Agent ServiceAppetite testing, referral and drafting orchestration.
Azure OpenAI modelsRisk narrative, pricing rationale and referral reasoning.
Azure AI SearchRetrieval over appetite guides, wordings and precedent.
Azure AI Content SafetyGuardrails on excluded factors and unsupported statements.
Azure API ManagementSecure gateway for policy, rating and broker platform APIs.
Azure SQLRisks, quotes, decisions and referral records.
Azure Cosmos DBSubmission state, versions and assessment history.
Azure Blob StorageSubmission packs, surveys and evidence artefacts.
Azure Key VaultPolicy system credentials and integration secrets.
Azure Communication ServicesQuote, referral and decline notifications to brokers.
Power BI / Microsoft FabricHit rate, appetite drift and portfolio dashboards.
04 — AI
What the agent consumes
The capabilities the copilot applies to every submission, and the line it does not cross.
AI capabilities embedded in the agent
- Document extraction
- Schedule and loss-run parsing
- Exposure enrichment
- Appetite matching
- Limit and exclusion checks
- Retrieval-augmented generation
- Risk summarisation
- Pricing rationale drafting
- Referral reason generation
- Fairness monitoring
- Portfolio analytics
- Workflow orchestration
AI responsibility boundaries
The copilot summarises, checks and drafts; a licensed underwriter prices, agrees terms and binds every risk. It does not decline a risk on its own, apply loadings outside the rating rules, or use factors excluded by policy or regulation. Each draft carries the extracted evidence, the appetite rule applied and the referral reason.
05 — Personalization
Personalization & evolving process
The same methodology applies to every agent in the catalog. Tune the line of business, the submission template, the appetite rules and the value model — the page structure stays identical.
Line & appetite profile
Define the lines of business, territories, limits, exclusions and the target portfolio mix. The personas are the broker, the underwriter, the referral authority and the portfolio manager.
Submission template
One consistent flow for every underwriting agent: receive, extract, enrich, assess against appetite, draft terms, refer where needed, decide and bind.
Appetite & authority rules
Declare the appetite statement, limit structure, exclusions, rating rules and the delegated authority matrix. Factors excluded by regulation are never used, even where correlated.
Value model
Capture baseline metrics first, then map the expected benefits: time to quote, submission capacity, hit rate, referral quality, data entry effort and audit evidence.
06 — Impact
Key outcomes & business impact
Starting targets for the value case — validate each one against the customer baseline during discovery.
- Time to quote−50%Submissions arrive decision-ready rather than as paperwork.
- Submission capacity2×More risks assessed without adding underwriting headcount.
- Referral qualityEvidencedEvery referral states its reason and attaches the evidence.
- Decision traceability100%Each decision stores the rule, rationale and underwriter.
Illustrative improvement index
Manual baseline = 100. Illustrative targets, not a commitment — confirm against the customer baseline.
- Time to quote: manual baseline 100, AI-assisted target 50
- Manual data entry: manual baseline 100, AI-assisted target 25
- Referral rework: manual baseline 100, AI-assisted target 45
Related & recommended
Derived automatically from our solution knowledge graph.
Related professional services
How we design, build and secure it.
ServiceAI Data Engineering
AI Data Engineering delivered by Cloud Mechanics certified experts.
ServiceMicrosoft Copilot Development
Microsoft Copilot Development delivered by Cloud Mechanics certified experts.
ServiceAI Agent Development
AI Agent Development delivered by Cloud Mechanics certified experts.
ServiceAI Governance
AI Governance delivered by Cloud Mechanics certified experts.
Related quick wins
Ready-made Azure AI to start fast.
Technologies
What powers this solution.
TechnologyAzure OpenAI
Enterprise access to GPT models with governance.
TechnologyMicrosoft Copilot Studio
Low-code platform to build custom copilots and agents.
TechnologyAzure AI
Managed AI services for vision, speech, language and document.
TechnologyAzure AI Foundry
Platform to design, evaluate and operate AI apps and agents.
Related managed services
Keep it running and optimised.
ManagedAI Managed Services
AI Managed Services from our UAE-based 24/7 Cloud Operations Center.
ManagedAzure Managed Services
Azure Managed Services from our UAE-based 24/7 Cloud Operations Center.
ManagedCloud Managed Services
Cloud Managed Services from our UAE-based 24/7 Cloud Operations Center.
ManagedDevOps Managed Services
DevOps Managed Services from our UAE-based 24/7 Cloud Operations Center.
Ready to move from challenge to solution?
Talk to a Cloud Mechanics expert or build your solution in minutes.