Commercial Underwriting at a European Insurer — 4× Quote Throughput on Mid-Market SME Risks
DocuMage + Multi-Agent Orchestrator + Coding Assistant pattern automating the data-capture and policy-rating layers of mid-market commercial underwriting.
The challenge
The client — a European commercial insurer specialising in mid-market SME commercial lines (property, business interruption, liability, marine cargo) with operations across several European markets — was losing market share to faster-quoting competitors. Average submission-to-quote turnaround was 5.8 working days for a typical mid-market commercial risk, against competitors quoting in 24-48 hours for comparable risk profiles. The broker channel — through which the bulk of the insurer's commercial business originated — was systematically routing time-sensitive risks to faster carriers.
The bottleneck was the underwriter workflow. A typical mid-market submission generated 60-80 pages of documentation: the broker's submission letter, the prospect's exposure schedules, prior-year loss runs, financial statements, the prospect's risk-management documentation, and the broker's narrative cover note. The underwriter was spending 70-80% of working time on data capture — extracting the key facts from this documentation into the rating engine's input schema, manually rating the risk, and assembling the quote letter — and 20-30% on the actual underwriting-judgement work the role exists for.
The constraints were significant. Solvency II governance requirements meant the underwriting decision had to be explainable and auditable. The insurer's existing rating engine could not be replaced (too much regulatory and operational integration depth) but the work feeding into it could be automated. The European multi-market footprint required the platform to handle multiple languages and the country-specific regulatory and commercial variations.
The approach
MindMap deployed DocuMage as the submission-document intelligence layer, Multi-Agent Orchestrator (Mo) as the submission-handling workflow coordinator, Coding Assistant-style pattern (with Summarization Wizard for the narrative components) for the quote-letter drafting, and Compliance Engine (Ce) for the Solvency II governance layer.
Phase one was the submission-intelligence layer. DocuMage was trained on five years of historical commercial submissions (with broker permission and PII redaction), learning to extract the key facts the underwriter needs: the insured entity's structure, the property and operations exposure, the prior-loss experience, the requested coverage limits and deductibles, and any specific risk-management features the broker has highlighted.
Phase two was the orchestration layer. For each submission, the orchestrator runs parallel extraction tasks across the submission document set, cross-validates the extractions (the entity name on the loss runs should match the entity on the financial statements; the requested limits should be consistent across the submission letter and the schedules), and produces a structured submission summary in the underwriter's review format. The rating engine is invoked with the extracted-and-validated data, and the quote-letter draft is generated by the summarisation layer.
Phase three was the underwriter workflow. The underwriter's review interface presents the submission summary, the rating-engine output, the suggested quote terms, the underwriter-judgement points that the system has surfaced (e.g. exposure features outside the insurer's standard rating tables, prior-loss patterns suggesting cover-condition tightening, broker-narrative themes worth probing) and the draft quote letter. The underwriter reviews and approves or modifies each component in the structured interface.
The pre-built building blocks
Rather than commission a ground-up build, the engagement leaned on MindMap's pre-built accelerator library — production-tested components that compress what would otherwise be a six-to-nine-month build into weeks.
DocuMage
Submission-document extraction across multi-broker template variants
Multi-Agent Orchestrator
Parallel submission-handling workflow and cross-validation
Summarization Wizard
Quote-letter narrative drafting
Compliance Engine
Solvency II governance and audit trail
The architecture
The platform runs on the insurer's Azure tenant in the EU region, with full data residency in-region and Solvency II-compliant audit trails on all underwriting decisions.
DocuMage runs a layered extraction model. The submission-document layout varies significantly by broker (each broker has their own submission template) and we handle this with broker-specific template recognition feeding into a broker-agnostic extraction model. The extraction model is a fine-tuned Llama 3.1 70B variant trained on the insurer's historical submissions, with structured-output constraints ensuring the extraction maps cleanly into the rating engine's input schema.
The orchestration layer uses Multi-Agent Orchestrator. Each submission triggers a graph of parallel tasks — entity extraction, exposure-schedule parsing, loss-run analysis, financial-statement spreading, coverage-request structuring, broker-narrative analysis. The orchestrator parallelises these and applies the cross-validation rules before invoking the rating engine and the quote-drafting layer.
Integration with the existing rating engine is via the engine's standard inbound API. The rating engine itself is unchanged; the platform feeds it cleanly-structured inputs faster and more consistently than the manual data-capture process did. Quote-letter generation uses the insurer's existing letter templates with the narrative components filled by Summarization Wizard from the underwriter-judgement summary.
Compliance Engine enforces the Solvency II governance overlay: every underwriting decision is logged with the extracted data, the rating-engine inputs and outputs, the underwriter-judgement points and resolutions, and the final quote terms with rationale. The audit trail is the evidence the insurer's internal audit and the regulatory inspector use for Solvency II reviews.
The numbers behind the story
Quote throughput per underwriter has risen approximately 4x. Average submission-to-quote turnaround has dropped from 5.8 working days to 1.4 working days for the in-scope mid-market commercial portfolio, with the bulk of the time saved on the data-capture layer the platform has automated.
Submission-to-quote auto-flow rate is 72% — meaning the platform takes the submission from receipt through to a draft quote without human touch on the data-capture and rating components, with the underwriter joining only at the review step. The remaining 28% of submissions involve enough novelty (unusual exposure features, complex multi-jurisdiction structures, specific broker requests) that the underwriter engages earlier in the flow.
Bind ratio — the proportion of quotes that result in bound business — has risen 31% on the in-scope portfolio. The improvement is attributable to two factors: the faster turnaround means the insurer is in-time on more time-sensitive risks, and the underwriter's freed time per submission has been redirected to broker-relationship work that has improved the insurer's broker-channel positioning.
Underwriting outcomes have not degraded. Twelve months of bound-business loss-ratio performance is statistically indistinguishable from the historical baseline, with the underwriter-judgement layer ensuring that the platform's data-capture acceleration does not translate into underwriting-discipline compromise.
The insurer's broker channel feedback has been the leading indicator: the insurer has moved from being characterised by brokers as 'slow but high-quality' to being characterised as 'quick and high-quality', which has translated into the insurer being on more broker shortlists for new business.
“We were losing time-sensitive risks to faster carriers because our submission-to-quote cycle was a structural problem. MindMap delivered four-times throughput in twenty-four weeks, without compromising underwriting discipline. Our brokers now characterise us as quick and high-quality rather than slow and high-quality — that is a strategic positioning change.”— Chief Underwriting Officer· European Commercial Insurer
Why MindMap was chosen
The insurer had evaluated three insurance-tech vendors. Two were US-headquartered and built on US-commercial-insurance assumptions that did not transfer cleanly to the European mid-market commercial context. The third was a European insurance-tech specialist whose document-intelligence depth was limited and whose multi-market regulatory handling was bolted-on rather than designed-in.
MindMap's DocuMage accelerator was already deployed in another European commercial-underwriting context, and we could demonstrate field-level extraction accuracy on the insurer's own sample submissions during the bid. The willingness to deploy entirely inside the insurer's Azure EU tenant with Solvency II-compliant audit trails was a critical commercial position.
Our embedded commercial-insurance domain expertise on the delivery team (two former commercial-underwriting heads from peer European insurers) was the third factor. The insurer's CUO felt that the team understood commercial underwriting, not just the data-capture automation.
Related deployments
Sovereign WhatsApp Banking
ChatNext-powered WhatsApp bot deployed inside the bank's air-gapped data centre, handling balance, transfers, statements and loan applications in English and Swahili.
Cheque OCR at 99.1% Accuracy
DocuMage replaced a legacy template OCR for cheque clearing, processing 10,000 cheques per day at 99.1% field accuracy with 94% straight-through processing.
UK Challenger Bank KYC
OnboardX rebuilt the KYC pipeline with liveness, sanctions and PEP enrichment in a single STP flow, collapsing onboarding from 5 days to 4 hours.
Want an outcome like this?
Start with a 2-week AI Readiness Sprint. We deliver a prioritised use-case backlog and business case grounded in what's actually buildable with our accelerator library.