Introduction

Moving from Expense Ratio to Loss Ratio – Unlocking value and potential

Where should the next wave of AI investment in specialty insurance be aimed?

In recent times, much of the insurance market’s investment in AI, automation and digital technology has focused on improving operational efficiency, with firms at varying stages of adoption.

That investment is starting to deliver value, but in the specialty lines business, many firms are still at an early stage in their adoption. As a result of this activity, processes are becoming faster, manual activity is being reduced and insurers are actively looking for opportunities to lower the cost of doing business.

In other words, much of the first wave of AI investment has been directed towards the expense ratio.

But is the next and potentially more valuable opportunity somewhere else? John Hele, a highly respected international insurance executive and now Executive Chair of Portage AI, argues that the next phase should focus much more directly on the loss ratio — the number underwriters are ultimately paid to manage. In his recent Best’s Review article, he makes the case that comparatively little of the first wave of insurance AI investment has been aimed at improving underwriting performance and loss outcomes.

The data already exists

The London Market already holds enormous volumes of valuable underwriting, claims and delegated authority information.

The challenge is that much of it remains difficult to use. And as early AI adoption matures, a clearer picture is emerging of what works. Generic large language models have proved impressive but not reliably accurate on specialty data, and their output is rarely auditable, so teams often end up checking everything anyway. Meanwhile, the long-term cost of owning and maintaining these systems remains unclear.

During this LMF community breakfast, John will talk through how submissions and claims information arrives in multiple formats. Different brokers, coverholders and cedants may describe the same data in different ways, while large-scale cleansing programmes can struggle to create and maintain a consistent view of historical information. Delegated authority and bordereaux data provide a particularly interesting example.

Once historic information is digitised accurately and mapped consistently, John argues that predictive models can identify relationships and patterns across policies, customers and claims that may previously have been difficult or impossible to see.

That potentially takes the AI conversation into a very different place.

Instead of asking only:
"How can AI make insurance operations more efficient?"
the Market can begin asking:
"How can AI help us make better underwriting, claims and portfolio decisions?"

Where could the value emerge?

Usable historical data can begin to influence three particularly important areas: claims leakage and fraud can potentially be identified earlier; underwriting requirements can be tested against actual loss experience rather than rules of thumb; and pricing can increasingly reflect the experience of an insurer’s own portfolio rather than relying solely on broader industry proxies.

That raises some important questions for underwriting and claims leaders. How close is the Market to being able to use AI in this way? What needs to happen to historic data first? How comfortable will boards, Lloyd’s and regulators be with information extracted using AI? And where might organisations see the earliest measurable impact?

Join the LMF discussion

At this LMF Community Roundtable Breakfast, we will bring together a small group of senior London Market practitioners to explore what the next chapter of AI investment could mean for specialty and commercial insurance.

The session will be facilitated by Roger Oldham, Founder & CEO of London Market Forums, and will combine a short presentation, live participant polling, Q&A and a frank and open roundtable discussion, and of course, the legendary LMF hot cooked breakfast.

This event will be particularly relevant to professionals with responsibility for underwriting, underwriting operations, claims, claims operations, delegated authority, portfolio management, pricing, exposure management, insurance operations and data strategy. All LMF events are FREE to access, and you don’t need to be a member to attend. The session is ONLY OPEN to insurance practitioners.

Around the table

We will explore three central questions:

1. After several years of technology investment, where are you seeing the benefit — the expense ratio, the loss ratio, or neither yet?
2. What do you feel your board, Lloyd’s or the Regulator need to see before relying on AI-extracted data to support a material insurance decision?
3. If your historical bordereaux and portfolio information were clean, consistent, and immediately usable tomorrow, what area of business would see the greatest benefit — claims, underwriting, pricing, exposure management, other?