Why the real AI advantage in insurance lies beyond the model

Senior executives are interesting not because they occupy the upper floors of an organisation, but because altitude brings perspective. 

Time spent making decisions across markets, portfolios and economic cycles produces a way of seeing the business that cannot easily be replicated by speed or technical proficiency alone. 

A junior analyst may manipulate a spreadsheet more deftly than the chief executive; few would suggest they therefore possess the same judgement.

If increasingly capable models become available to everyone, then the source of competitive advantage begins to shift elsewhere. It lies less in the intelligence itself than in the environment within which that intelligence operates.

Ben Rose, Supercede

Artificial intelligence finds itself in a surprisingly similar position.

The wrong question

Much of the discussion surrounding AI in insurance continues to revolve around models. Which is best? Which is safest? Which is improving fastest? The answers change with remarkable frequency, and will continue to do so. 

What was state of the art six months ago already feels routine. The pace of progress is such that today’s competitive advantage has an awkward habit of becoming tomorrow’s minimum expectation.

This has an interesting consequence. If increasingly capable models become available to everyone, then the source of competitive advantage begins to shift elsewhere. It lies less in the intelligence itself than in the environment within which that intelligence operates.

Context over content

Insurance has always been a business in which context matters more than individual facts. 

A renewal is meaningful because of the years that preceded it. A placement derives significance from the portfolio it supports. A relationship between broker and carrier often explains more than the wording of a single contract. 

Much of the industry’s knowledge exists not within documents, but between them, woven through workflows, accumulated experience and the institutional memory of the people involved.

That presents AI with a rather different challenge from the one implied by benchmark scores and model comparisons. Reading a document is relatively easy. Understanding why that document exists, how it came to be, what preceded it and what depends upon it is considerably harder. Yet that is precisely the difference between processing information and exercising judgement.

Value

The organisations that derive the greatest value from AI are therefore unlikely to be those with exclusive access to a particular model. They are more likely to be those that succeed in giving whichever model they choose the richest possible representation of their business. 

Trusted data matters because unreliable information is difficult for humans and machines alike to recognise. Connected workflows matter because they reveal how decisions emerge rather than simply recording their outcomes. Shared context matters because it provides the continuity from which sound judgement is formed.

This is why infrastructure deserves more attention than it often receives. As the industry moves towards agentic AI, model choice will become increasingly flexible. Organisations will experiment, replace, upgrade and combine models with far greater ease than they replace the systems that provide those models with their understanding of the business. 

The enduring asset is therefore unlikely to be the intelligence itself. It is the vantage point from which that intelligence observes the market.

That observation sits at the heart of Supercede’s approach to reinsurance intelligence. The ambition is not to persuade the market that one foundation model will permanently outperform another. History suggests no such advantage lasts for long. 

Instead, the objective is to build a connected representation of reinsurance - one founded on trusted data, structured workflows and shared institutional memory - so that whichever intelligence an organisation chooses can develop a deeper understanding of the market it serves.