Insurance Policy Platform
We keep coming back to one thesis: a lot of legacy enterprise software is protected by a moat that no longer exists. The license fee is big, the switching cost is real, but the underlying product is rebuildable on a modern foundation in a fraction of the time it used to take.
So we picked a category where that thesis is uncomfortable to argue with, insurance policy administration, and stood a product up against a well-known legacy system carrying a seven-figure annual license. A few weeks of focused build time got us to honest feature parity for an agency, end to end.
What we built
We built the shape an actual carrier or MGA would need, not a marketing demo:
- Public marketing and landing pages.
- Dynamic quote flows that branch by product and jurisdiction.
- Online payment collection.
- Account creation and customer sign-up.
- Document generation for policies, declarations, and notices.
- Full policy management across the lifecycle.
- Multi-tenant configuration, from a single agency to an MGA-style tenant with hundreds of brokers, each carrying different products, all under one tenant.
The same surfaces a real customer touches, and the same surfaces an admin, broker, or operator lives in.
Why it was possible in a few weeks
The honest answer is that the slow, expensive parts were already done before the build started.
- A modern stack, already wired. TypeScript end to end, a current framework, a typed API client, a managed database, deployment, and observability, on a foundation an AI assistant can actually reason about.
- A design system, implemented in code. Tokens, primitives, forms, tables, navigation, dashboards. Production-ready, with the small details right: tabular numerals, hairline borders that respect dark mode, accessible focus.
- A headless API at the core. The web app is one consumer of the backend, not the backend itself.
That meant the build time went almost entirely into insurance behavior: rating, eligibility, quote shaping, policy state, document templates, payment hooks, broker hierarchies, tenant configuration, instead of rebuilding the product surface from scratch.
Multi-tenant was the real test
A single-product demo for a single agency is not parity with a legacy policy- admin system. The legacy moat is partly about how messy this industry actually is, so the build had to handle the real shape: one small agency with their own branding and forms, and an MGA-style tenant managing hundreds of brokers with different products, commissions, and document templates, across jurisdictions, behind role-based access.
Almost none of that needed new "platform" code. It needed disciplined data modeling, a clean permission boundary, and components that could be configured rather than rewritten per tenant.
Why it matters
A few weeks to feature parity is the headline, not the lesson. The lesson is that the parts of an enterprise build that used to be hard (design system, headless API, modern tenancy, AI-assisted implementation) are now infrastructure. Invest once and they compound. Categories that looked untouchable become addressable by small, focused teams.
This was the test of that idea. It held up, and it's the same playbook we bring to client work.