Client overview
- Industry focus
- Insurance
- Portfolio segment
- Finance
- Organization profile
- Mid-market P&C insurer, ~850 employees, multi-state footprint
The carrier had grown through acquisition; each book of business carried its own policy administration quirks, spreadsheet bridges, and broker-facing portals. Underwriting teams rotated between three legacy consoles to complete a single multi-line quote, and claims adjusters replayed steps in spreadsheets when core systems disagreed on coverage codes. Regulatory reporting required month-end reconciliation that pulled four staff away from higher-value adjudication work. Leadership needed a unified experience that could absorb future books without replatforming again — while proving Sarbanes-Oxley–grade controls on financial postings tied to premiums and refunds.
Problem
Fragmented legacy tools slowed policy changes and introduced inconsistent data across brokers, underwriters, and claims teams.
Operational reality was a patchwork of green-screen edits, emailed endorsements, and BI extracts that disagreed after midnight batch windows. Brokers saw stale eligibility flags until the next sync, causing quoting rework and erosion of placement confidence. Claims supervisors lacked a single timeline showing reserve movements, SIU referrals, and external counsel documents in one audit trail.
Pricing models lived partly in spreadsheets because actuarial APIs could not version alongside rate filings in flight. QA cycles were manual and seasonal; regression suites took four days each release, forcing the business to batch risky changes instead of shipping incremental improvements.
Compliance teams flagged that immutable logs for sensitive fields (limits, deductibles, beneficiary updates) were incomplete when edits traversed asynchronous queues. Incident retrospectives revealed three near-miss PII disclosures tied to attachment handling in legacy workflow engines.
Executive sponsors framed the mandate clearly: consolidate policy lifecycle and claims workflow behind role-based experiences, integrate with downstream finance for earned premium recognition, and instrument every critical path for observability — without freezing new product launches already tied to reinsurance treaties.
Solution
We engineered a modular web platform with role-based workflows, audit trails, and integrations to core administration systems. QA automation guarded critical pricing and eligibility paths before every release.
We anchored the architecture on a domain-driven design map separating quoting, issuance, billing alignment, and claims servicing contexts. Each bounded context exposes REST contracts versioned alongside OpenAPI schemas checked into CI; GraphQL was intentionally avoided at the edge to keep cache semantics predictable for regulatory snapshots.
The customer-facing tier uses Next.js with server components for document-heavy screens and client islands for interactive rating grids. Session handling flows through an OAuth2 broker integrated with the insurer’s IdP; fine-grained authorization combines RBAC with attribute checks on jurisdiction and product line.
Claims journeys orchestrate tasks through a durable workflow engine backed by PostgreSQL with optimistic concurrency on financial postings. Attachments route through virus scanning and object storage with signed URLs and virus rescan on litigation holds. Event streams feed an analytics warehouse for loss triangles without blocking OLTP commitments.
Playwright suites model producer, underwriter, and adjuster personas against ephemeral environments seeded from anonymized fixtures. Contract tests validate integrations to policy admin and payment gateways using stubbed providers in Docker Compose pipelines.
Implementation
- 1
Discovery & regulatory alignment
Six-week inception mapped AS-IS journeys, identified systems of record, and produced a filing-safe release plan. We paired with compliance to tag data classes, define retention, and draft control matrices for SOC2-style evidence. Parallel tracks profiled claims payment latency and broker NPS deltas by state.
- 2
Vertical slice delivery
The first slice covered personal auto renewals end-to-end: quote, bind, endorse, FNOL intake, and cash application. Feature flags guarded partial rollouts by state. Data migration scripts ran in dry-run mode for two close cycles before cutover weekends with automated reconciliation reports.
- 3
Hardening & scale validation
We executed load tests on claims document peaks, chaos exercises on broker API dependencies, and game-day failover drills for RDS multi-AZ. Observability dashboards linked SLO burn rates to incident playbooks; on-call rotations trained on synthetic claim seeds before peak season.
Tools & platforms
- GitHub Actions for build, test, and blue/green ECS deploys; Snyk on PRs; OWASP dependency track for third-party libs.
- Terraform modules for VPC, RDS, Secrets Manager, and WAF rules; VPC endpoints to keep traffic off public internet.
- LaunchDarkly-style feature flags implemented with environment-backed toggles and kill switches for pricing engines.
Engineering challenges addressed
- Harmonizing historical policy numbers and endorsements across acquired systems without blocking open claims.
- Keeping sub-200ms p95 read paths for broker dashboards during month-end extracts.
- Coordinating zero-downtime cutover while parallel financial subledgers reconciled earned premium.
Program artifacts & environments


Tech stack
- Next.js
- TypeScript
- Node.js
- PostgreSQL
- Playwright
- Docker
- AWS
- OpenAPI
- Terraform
Results
- ~40% reduction in processing time for standard claims
- Near-zero regression defects on billing-critical flows post-launch
- Audit-ready history for compliance reviews
Quantified impact
39% faster cycle time for straightforward claims through first payment
Measured over 180 days vs. baseline; controlled for seasonality.
78% reduction in broker-reported quote errors tied to stale eligibility
Based on ticketing volume and sampled QA of bound policies.
~$1.1M annualized ops efficiency from reduced manual reconciliation FTE
Blended rate model; excludes one-time migration costs.
Key takeaways
- Regulated insurance modernizations succeed when pricing, claims money movement, and audit lineage are modeled as first-class workflows — not bolted spreadsheets.
- Investing early in persona-based test data and parallel run strategies de-risks cutover more than big-bang UAT cycles.
- Thin, versioned integration contracts beat all-in-one APIs when multiple core systems must coexist during multi-year migrations.
