InsurifyAI
AI Insurance Claims Platform
AI-powered insurance claims platform with a Liquid Glass design system across 33 dashboard pages, 21 autonomous Shadow Team agents handling feature development, real-time voice coaching during claims calls, automated demand letter generation, and certified-mail delivery via Lob. Founded on a real Tesla settlement that grew from $17K to $29K — a +71% recovery — through 200 pages of legal evidence.
By the Numbers
Core Features
InsurifyAI fuses insurance domain expertise with full-stack engineering and a fleet of autonomous AI agents. Each capability was built to win specific carrier-disputed claims — not as a hypothetical feature list.
Real-Time Voice Coaching
Live transcript + AI-suggested rebuttals during recorded claims calls. Coach insureds through adjuster calls in real time using Claude + WebRTC streaming.
Demand Letter Generation
python-docx pipeline produces court-ready document layouts with exhibit numbering, citation formatting, and bates stamps. Each letter cross-references NADA + KBB + Manheim valuations and applicable state statutes.
Shadow Team Autonomous Agents
21 AI agents handle the feature roadmap: PR routing by tier, automated testing, review escalation, and merge gates. Each agent operates within budget guardrails.
Certified Mail Delivery
Lob API integration handles USPS Certified Mail dispatch. Letters print, envelope, and ship with tracking numbers captured automatically as proof of delivery for the claim record.
Liquid Glass Design System
Apple HIG-inspired UI with translucent panels and depth-aware components shipped across all 33 dashboard pages. Built on shadcn/ui with custom material overlays.
State-Specific Legal Database
50-state legal database with bad-faith statutes, statute of limitations, and CCC ONE valuation methodology. Letters cite the specific applicable statute per claim location.
Tesla Settlement Demand — $17K → $29K
The real-world legal claim that drove the platform: a 2019 Tesla Model 3 case involving FSD, CCC ONE valuation disputes, and a Feb 2025 incident. State Farm's initial offer was $17,000. Through a 200-page legal demand packet delivered via USPS Certified Mail, the State Farm Claims Upload Center, and email, the settlement was negotiated to $29,000 — a +71% increase, $12K more than the initial offer. Every workflow now automated by InsurifyAI started here.
Why this approach worked: insurance domain expertise (50-state legal database, CCC ONE valuation methodology, NH bad-faith statutes), technical precision (python-docx formatter for court-ready layouts), cross-referenced evidence (NADA + KBB + Manheim comparables, FSD-specific damage docs, policy language analysis), and the full stack — engineering to build the tools, business to negotiate the leverage, insurance domain to identify the carrier's weak points.
Technical Architecture
InsurifyAI runs on AWS Fargate with 5 ECS services, 123 Terraform-managed resources, a Supabase Auth + Postgres backend (pgvector for legal-document search), and a Next.js App Router frontend. Shadow Team agents operate against the same data plane with tiered PR routing.
Backend
- 5 ECS Fargate services (API, Web, Realtime, Agent-Hub, Agent-Runtime)
- RDS PostgreSQL 16 with pgvector
- ElastiCache Redis for queues + session
- Supabase Auth (SSR) shared across the suite
- Stripe webhooks for billing events
Frontend
- Next.js App Router (33 pages)
- Tailwind CSS + shadcn/ui
- Liquid Glass design system (Apple HIG)
- CloudFront CDN distribution
- WAF rate-limit + bot challenge
AI & Agents
- Claude API for letter generation + voice coaching
- 21 Shadow Team agents (tiered PR routing)
- Agent-hub orchestrator with budget guardrails
- 50-state legal citation database
- python-docx court-ready document layout
Infrastructure
- Terraform-managed VPC + subnets + IAM
- 123 cloud resources across 4 envs
- Cloudflare Email Routing (28 addresses)
- SendGrid transactional relay
- GitHub Actions CI/CD with PR previews