Why MRC slips defeat generic document AI

The Market Reform Contract is the London Market’s standard placing document — and a stress test that generic document AI reliably fails. An MRC is not a form: it is a structured narrative whose sections, risk details, conditions and schedules interrelate in ways that matter commercially.

Where horizontal tools break

Tools trained on invoices and application forms treat an MRC as text to be mined. They extract words but lose structure: conditions detach from the sections they modify, schedule references break, multi-class slips blur together. The output looks like data and reads like noise — which is worse than nothing, because it must be checked line by line.

What reliable MRC extraction requires

Reliable extraction treats the MRC’s structure as first-class: sections recognised as sections, schedules linked to risk details, terms tied to the classes they govern. It also requires market context — knowing what a line, an order, or a TIV split means — and validation against the rest of the submission pack. That is domain software, not document tooling; it is why London Market extraction remains a specialist discipline.

Keep reading

What is insurance submission ingestion? A practical guide · The AI underwriting workflow: what changes, what doesn’t · All articles