• contact@verticalserve.com
Home / Engineering / Post 11
Engineering Blog · Post #11

One-Click AI Analysis: 8 Parallel Intelligence Checks on Every Submission (Extract, Enrich, STP, Appetite, Rate, Similar, Assign, CAT)

How InsightUW runs 8 AI-powered intelligence checks in parallel on a D&O Public Company submission — streaming results to progressive UI cards in under 12 seconds — so the underwriter sees a fully analyzed risk before they finish reading the cover email.


The Problem

Underwriters are drowning in manual pre-analysis. Before they can even form an opinion on a risk, they must:

  1. Extract key data points from the application (insured name, revenue, industry, limits requested)
  2. Enrich the submission with external data (D&B firmographics, SEC filings, litigation history)
  3. Check appetite against the company's underwriting guidelines
  4. Run a rate indication against the pricing model
  5. Search for similar risks in the book for benchmarking
  6. Evaluate STP eligibility — can this go straight through?
  7. Determine assignment — which underwriter should handle this?
  8. Run CAT exposure models for natural catastrophe accumulation

Each of these steps involves a different system, a different screen, and a different manual process. An experienced underwriter spends 20–45 minutes on this pre-analysis per submission. A junior underwriter takes longer and misses things.

The real cost is not just time — it is the sequential bottleneck. These 8 checks are largely independent of each other. There is no reason they cannot run simultaneously.

The InsightUW Approach

InsightUW's AI Intelligence Panel runs all 8 checks in parallel the moment an underwriter clicks a single "Analyze" button — or automatically on submission creation if configured. Results stream to the UI in real-time using Server-Sent Events (SSE), rendering as progressive cards that fill in as each check completes.

graph TB subgraph Trigger["Analysis Trigger"] A["One-Click 'Analyze' Button"] B["Auto-Trigger on<br/>Submission Create"] end subgraph Orchestrator["Parallel Orchestrator"] C["Analysis Job Manager<br/>8 Parallel Workers"] end subgraph Checks["8 Intelligence Checks (Parallel)"] D1["1. Extract<br/>AI Data Extraction"] D2["2. Enrich<br/>External Data Enrichment"] D3["3. STP<br/>Straight-Through Processing"] D4["4. Appetite<br/>Guidelines Check"] D5["5. Rate<br/>Pricing Indication"] D6["6. Similar<br/>Book Search"] D7["7. Assign<br/>Underwriter Matching"] D8["8. CAT<br/>Catastrophe Exposure"] end subgraph Stream["SSE Result Stream"] E["Progressive Card Rendering"] F["8 Cards: Loading → Complete"] end A --> C B --> C C --> D1 C --> D2 C --> D3 C --> D4 C --> D5 C --> D6 C --> D7 C --> D8 D1 -->|"SSE"| E D2 -->|"SSE"| E D3 -->|"SSE"| E D4 -->|"SSE"| E D5 -->|"SSE"| E D6 -->|"SSE"| E D7 -->|"SSE"| E D8 -->|"SSE"| E E --> F

Check 1: Extract (AI Data Extraction)

Parses the submission documents (ACORD app, cover letter, attachments) using LLM-based extraction to pull structured fields.

Check 2: Enrich (External Data Enrichment)

Pulls firmographic, financial, litigation, and regulatory data from external providers.

Check 3: STP (Straight-Through Processing Evaluation)

Determines if the submission qualifies for automatic processing without underwriter intervention.

Check 4: Appetite (Underwriting Guidelines Check)

Evaluates the submission against the carrier's appetite guidelines for the specific LOB, industry, and limit.

Check 5: Rate (Pricing Indication)

Runs the submission data through the carrier's pricing model to generate a preliminary rate indication.

Check 6: Similar (Book Search)

Searches the carrier's book of business for similar risks to provide benchmarking context.

Check 7: Assign (Underwriter Matching)

Determines the best underwriter for this submission based on expertise, workload, and authority level.

Check 8: CAT (Catastrophe Exposure)

Evaluates natural catastrophe accumulation impact. For D&O this is typically low, but the check still runs for completeness and for LOBs where it is critical (Property, Builders Risk, Marine).

Progressive Card Rendering via SSE

The 8 checks stream their results to the frontend via Server-Sent Events. The UI renders 8 cards, each starting in a "loading" state and transitioning to "complete" as results arrive.

sequenceDiagram participant UW as Underwriter participant UI as InsightUW UI participant API as Analysis Orchestrator participant W1 as Extract Worker participant W2 as Enrich Worker participant W3 as STP Worker participant W4 as Appetite Worker participant W5 as Rate Worker participant W6 as Similar Worker participant W7 as Assign Worker participant W8 as CAT Worker UW->>UI: Clicks "Analyze" UI->>API: Post /api/v1/submissions/SUB-2026-04-1847/analyze API-->>UI: SSE stream opened (analysis id: AN-20260421-001) par 8 Parallel Checks API->>W1: Extract API->>W2: Enrich API->>W3: STP API->>W4: Appetite API->>W5: Rate API->>W6: Similar API->>W7: Assign API->>W8: CAT end Note over UI: All 8 cards show spinner W7-->>API: Assign complete (640ms) API-->>UI: SSE: {"check":"Assign","status":"Completed"} Note over UI: Assign card renders ✅ W3-->>API: STP complete (890ms) API-->>UI: SSE: {"check":"STP","status":"Completed"} Note over UI: STP card renders ✅ W8-->>API: CAT complete (1100ms) API-->>UI: SSE: {"check":"CAT","status":"Completed"} W4-->>API: Appetite complete (1560ms) API-->>UI: SSE: {"check":"Appetite","status":"Completed"} W6-->>API: Similar complete (1870ms) API-->>UI: SSE: {"check":"Similar","status":"Completed"} W1-->>API: Extract complete (2340ms) API-->>UI: SSE: {"check":"Extract","status":"Completed"} W5-->>API: Rate complete (3200ms) API-->>UI: SSE: {"check":"Rate","status":"Completed"} W2-->>API: Enrich complete (4120ms) API-->>UI: SSE: {"check":"Enrich","status":"Completed"} Note over UI: All 8 cards complete ✅ Total: 4.1 seconds

Use Case: D&O Public Company — Solaris Energy Holdings

The Scenario

An underwriter at Global Re opens submission SUB-2026-04-1847 for Solaris Energy Holdings — a $4.5B revenue public solar energy company requesting $20M D&O Side A/B/C coverage. Instead of spending 35 minutes pulling data from four different systems, the underwriter clicks a single button.

What Happens in InsightUW (Timeline)

Time Event System Action
10:00:00 AM Underwriter clicks "Analyze" POST to analysis orchestrator; SSE stream opens
10:00:00 AM 8 parallel workers launched Extract, Enrich, STP, Appetite, Rate, Similar, Assign, CAT
10:00:00.6s Assign card completes Patricia Reeves recommended (expertise 0.94, workload 14/20)
10:00:00.9s STP card completes Not eligible — $20M limit exceeds STP max, active litigation
10:00:01.1s CAT card completes No direct CAT exposure; market downturn correlation noted
10:00:01.6s Appetite card completes Within appetite (score 78); ESG risk warning; senior UW referral required
10:00:01.9s Similar Risks card completes 4 similar risks found; benchmark premium $401K avg
10:00:02.3s Extract card completes 18 fields extracted with 96% confidence
10:00:03.2s Rate card completes Indication premium $485K; range $425K–$560K
10:00:04.1s Enrich card completes D&B 5A2, 1 active derivative suit, ESG score 72, no restatements
10:00:04.1s All 8 cards rendered Full intelligence panel visible. Total elapsed: 4.1 seconds
10:00:04.1s Underwriter begins review All pre-analysis complete. Decision context ready.

The underwriter now has everything they need to form an opinion — extracted data, external enrichment, appetite validation, a rate indication, similar risk benchmarks, assignment recommendation, STP determination, and CAT exposure — all in 4.1 seconds instead of 35 minutes.

Metrics: Before and After InsightUW AI Analysis

Metric Before InsightUW After InsightUW Improvement
Pre-analysis time per submission 20–45 minutes 4–12 seconds 99.5% faster
Systems accessed for pre-analysis 4–6 separate applications 1 click, 1 panel 83% fewer systems
Data extraction accuracy 78% (manual entry) 96% (AI extraction) 18% improvement
STP identification rate 5% (manual identification) 22% (automated scoring) 4.4x improvement
Rate indication availability After full review (days) Immediate (seconds) From days to seconds
Similar risk benchmarking Rarely done (too time-consuming) Every submission, automatically 100% coverage
Underwriter capacity (submissions/day) 6–8 12–16 2x throughput
Time to first underwriter action 2–4 hours < 5 minutes 97% faster

Key Takeaways

  1. Parallelism is the key insight. The 8 intelligence checks are largely independent. Running them sequentially is a design failure, not a technical necessity.

  2. SSE streaming makes the UI feel alive. Progressive card rendering gives the underwriter immediate feedback and lets them start reviewing the first results while later checks are still running.

  3. One click replaces 35 minutes of manual work. The underwriter's time is redirected from data gathering to decision making — where their expertise actually adds value.

  4. Every check produces structured, auditable output. The analysis results are stored with the submission, creating a full audit trail of what the AI found and recommended.

  5. STP identification at scale transforms throughput. Automatically identifying the 20–25% of submissions that can go straight through processing doubles effective underwriter capacity.


Ready to give your underwriters 8 checks in one click? InsightUW's AI Intelligence Panel transforms pre-analysis from a 35-minute manual process into a 4-second automated workflow.

Schedule an AI Analysis Demo →

See InsightUW run on your data

A 45-minute working session with a real broker email and your LOBs.

Request a demo