TL;DR

MS Amlin expects an AI system its staff call Wordy Bot to halve the time it spends checking the language of reinsurance contracts, Insurance Business UK reported on 30 October. The tool compares contracts with the insurer’s underwriting standards and flags gaps for human review, and it already covers North American property reinsurance as the 1 January renewal season approaches. The time saving is MS Amlin’s own expectation, not a measured result.

What the tool does

Wordy Bot reads a draft reinsurance contract and highlights missing clauses, departures from standard wording and possible inconsistencies, which underwriters and the wordings team then assess. The insurer expects less back-and-forth with brokers as a result, so business gets placed sooner. It also aims to show underwriters how a change in wording shifts coverage and exposure, both for a single contract and across the portfolio.

Chief underwriting officer Martin Burke was clear that the tool makes no underwriting decisions. In his account it moves underwriters’ time from finding wording changes to interpreting them, and gives brokers “the potential for quicker reviews, fewer avoidable questions and greater clarity over the wording before a contract is finalised”.

Teams in London and at group company MSIG Asia in Singapore built it under an automation drive run across the MSIG group. The project lead was Dr Fei Ling Woon, MS Amlin’s head of robotics. Casualty reinsurance in North America is next in line, followed by its other reinsurance lines and, in time, direct insurance.

Why wordings matter now

Insurance Business argues the timing matters because reinsurance markets are softening. With cedants pressing for broader terms, minor edits slipped into renewal paperwork can broaden cover unnoticed, so spotting them carries more weight than when reinsurers dictate terms. The flip side for brokers is that changes are less likely to slip through under deadline pressure.

Looking Forward

The cautious design matches the sector’s mood. ISG research published in October, cited in the report, found three-quarters (75%) of insurance decision-makers would rely only on models built for insurance, or on AI paired with governed human review, when deciding underwriting and claims. Only 6% were content with a general-purpose model working alone. In our view, contract checking is the kind of contained, document-heavy task where London’s insurance market can show AI returns without handing over judgement, a theme running through recent deals such as Guidewire’s purchase of hyperexponential.