Connecting a broker's back office to eight insurers at once

The broker placed complex commercial risks — fleets, warehousing, professional indemnity — across a panel of eight insurers. Each insurer had its own portal, its own field names and its own idea of what a postcode looked like, so a single quote meant typing the same forty fields eight times. Account handlers spent more of their day copying data than advising clients, errors crept in between portals, and a client asking for comparison terms routinely waited two working days. The broker's policy administration system was twenty years old and had no API of its own.

1. 01

   Mapped every insurer's data model against the broker's own before writing any code. Most of the eventual complexity lived in the differences — eleven ways of describing a building's construction, three of them contradictory.

2. 02

   Designed a single canonical risk model and an adapter per insurer, so adding the ninth insurer is a contained piece of work rather than a change to everything.

3. 03

   Used insurer APIs where they existed and carefully rate-limited portal automation where they didn't, with every submission logged so a handler can see exactly what was sent where.

4. 04

   Read from the legacy policy system through a narrow, read-only integration instead of replacing it — the broker needed faster quoting, not a risky core-system migration.

5. 05

   Ran it in parallel with the manual process for a month, comparing every automated submission against the hand-keyed one until the handlers trusted it.

An account handler now enters a risk once and has comparable terms from the whole panel back the same afternoon. Re-keying errors between portals have disappeared because there is only one set of data. The team handles more renewals per head without new hires, and the audit trail of every submission has become the first thing the compliance lead opens when an insurer queries a placement.

• TypeScript
• Node.js
• PostgreSQL
• Playwright
• Azure
• REST & SOAP