Cytora and Relativity6 see Beazley experience as a model partnership 

Relativity6’s industry classification modelling is enhancing the underwriting process on Cytora’s digital risk processing platform, the chief operating officers at the two companies say

Insurtech Cytora and data firm Relativity6 plan to replicate the partnership they formed at Beazley in the US with other carriers and in other geographies, according to their chief operating officers (COOs), Juan de Castro and Josh Lurie.

Cytora is using the industry classification modelling of Relativity6 to enhance the underwriting process on its digital risk processing platform.

In an interview with Insurance Day, de Castro and Lurie say their integration enables underwriters to accelerate commercial quotes while using more accurate risk insights with intelligent industry classification data. The process turns data from freeform text into standard industry classifications.

Cytora says its platform enables insurers to digitise underwriting workflows from submission to quote, “operationalising” different sources of data across their lines of business, including for risk clearance, onboarding and triage. Operationalising data means ensuring it is used consistently across processes and that the outcomes are consistent.

De Castro says: “With Relativity6 data, we have done two things. One is we attach it to every single risk, so a human doesn’t need to do that manually. That means when an underwriter receives a risk, it already contains the data. The second thing is two underwriters might make different decisions, so we’re driving consistency of outcomes based on that same data.”

Its partnership with Relativity6 follows the launch of the latest enhancement to Cytora’s platform: leveraging large language models alongside its proprietary artificial intelligence (AI).

Their partnership began at, and thanks to, Beazley. De Castro says: “A year ago, we met through a common client, which is the best way of starting a partnership. Beazley was about to begin working with Relativity6 and Cytora separately and they had done their due diligence on our companies.”

“It started with Beazley in the US,” Lurie continues, “and we’re already talking about expansion. We have a success story together, so the plan is to look at our collective clients and prospects and have more success stories. That will happen over this year and next.”

In the medium term, they see two potential angles to expanding their partnership: by adding workflows within the same insurer, such as the renewals process, and by adding other types of data points required for underwriting.

Improved accuracy

The thrust of their joint offering is the use of AI to improve the accuracy and speed of an underwriter’s understanding of an insured.

Formerly COO at Hiscox, de Castro quotes the specialty insurer’s eponymous founder and chairman. “Robert always said: ‘You need to ask three things to underwrite a risk: who is the client, who is the client and who is the client?’ But what happens without Relativity6 is underwriters don’t have an effective way of matching a client with a given classification code.”

Lurie points out most of the 33 million or so companies in the US have fewer than 20 employees.

“Historically, insurers depend on outdated government filings from the moment these small businesses are incorporated and, as we know, in the small business and micro-business world, companies are constantly changing; they’re going in and out of business. Ironically, then, one of the most important pieces of information is essentially outdated the second the underwriter touches it.”

Relativity6 has an “end-to-end AI solution” that finds, classifies and monitors businesses and their activities, Lurie says. It searches the entire open public web to determine all of this in “under two seconds”. More than that, it forms an “original opinion” of each insured and delivers this through different classification taxonomies, such as National Association of Insurance Commissioners code and Standard Industrial Classification codes.

Relativity6 includes with every result it delivers on an insured a “confidence score”.

Lurie says: “We spend a lot of time with our clients understanding the importance of thresholding against this confidence score, which is critical. This is modernising the classification function in the pre-fill workflow for underwriters and speed is a critical element of that, especially when you’re dealing with small businesses where insurance companies are writing premiums that are not huge.”

He continues: “It means underwriters can apply the least amount of touch and at the same time get the most value through their workflows.”

Three levers of modernisation

De Castro sees market modernisation as using technology to reimagine insurance processes and drive value to clients, brokers and shareholders through three main levers.

These are: greater efficiency, by removing low-value activities so underwriters can instead focus on those where judgment is required; an improved broker service, whereby insurers can be more responsive to brokers; and leveraging data better to drive enhanced risk selection.

“If you think through these three goals, you’re driving a lower expense ratio by increasing efficiency and also a lower loss ratio through improved or enhanced risk selection,” de Castro says. “This results in value to shareholders and to clients because, in a competitive market, insurance products will become more affordable.”

Cytora has enabled its clients to double the productivity of their underwriters, namely, the amount of premium they write, de Castro says. It has also enabled them to “massively shorten” their responsiveness to brokers, from several days, to a few hours.

Accuracy in pricing risk, Lurie adds, is essential to reduce premium leakage, which is a “multibillion-dollar problem”. He says: “When an insurance company is writing a policy for, let’s say, a carpenter, but they find out after the fact it was a roofer, the miscalculation on the premium they’re charging is the premium leakage. And then, if that person falls off a roof, that’s going to start impacting the loss ratio.”

Human on the loop

The common theme of all modernisation efforts, de Castro continues, is a transition from what is typically described as having a human in the loop to having a human on the loop.

“In the loop is where every step of every process requires human intervention, while on the loop is where technology is performing all the lower-skilled activities and human intervention is only deployed where and when it matters,” he says.

In its Global Risks Report 2024, the World Economic Forum had the title “AI in charge” in the section describing risk from the perceived potential adverse impacts of AI. De Castro says the notion of “AI in charge” stems from fears of, for example, AI hallucination, which is the term for an incorrect or misleading result an AI model generates.

“You want AI to be supporting the underwriter,” de Castro says, “but the main segments of commercial insurance are too complex for AI and we are far away from AI doing the whole of underwriting.”

He adds AI is “not very good at predicting the future”, so the judgment of an underwriter on risk, such as from climate change and socio-economic changes, continues to be required.

“You want AI to be supporting the underwriter but the main segments of commercial insurance are too complex for AI and we are far away from AI doing the whole of underwriting” Juan de Castro, Cytora

Lurie stresses the need for purpose-built AI solutions. “That is where the most efficiency and effectiveness is going to come from, versus these grandiose, broader concepts of AI to totally disrupt the underwriting processes and risk models that have been around for many years.”

The main risk from AI, he continues, is the rapid pace of its development, meaning there will be many options for insurance companies to choose from as they “orchestrate” the use of AI models.

“Ironically, the goal of greater efficiency could create an inefficient process of trying to figure out how to use all this new technology,” Lurie says, “because this is such an old industry that is set in its ways in a lot of respects.”

He cautions against claiming AI is of itself a risk when instead the crux is figuring out how best to use it. “We have an opportunity now to use purpose-built AI solutions and really focus on specific issues and problems. If we can do that then the efficiencies we’ll see in workflows are going to be amazing.”

On how Cytora and Relativelty6 protect their AI technology from cyber attack, they highlight the International Organization for Standardization (ISO 27001) and Solvency II as among the “critical elements”.

These requirements are now “front and centre” and part of a company’s infrastructure “right out of the gate”, Lurie stresses. “Relativity6 addressed security concerns way earlier in our evolution than most technology companies I saw in the past did. When you’re servicing insurance and financial services, industries that are hypersensitive to security risks, you must go above and beyond, which is why we protect the models we deliver to clients through the required international standards.”

De Castro highlights the “exceptionally high security measures” in place at re/insurance companies before the advent of AI as a serious tool for their sector. “Cytora was built as ISO 272001 certified from the very beginning, because we are handling customer data,” he says.

“We go very deep in terms of classifying each data point of every request we’re receiving and we have several layers of encryption in our architecture. AI doesn’t really change the security measures you need to protect your data because you should have been protecting it already,” he adds.

Across classes of business

Their approach can be applied to every product line because these have the common issue of manual underwriting workflows. “The data itself will vary across lines of business, but in every single one of them underwriters use external data to analyse the risk,” de Castro says.

Lurie adds Relativity6 is focused on property/casualty and any product lines within that class of business, but is exploring group health as a possible additional area for its modelling.

On interconnected and systemic risk, de Castro says Cytora is working with its clients on aggregation. “Aggregation within a line of business is pretty well managed and understood, but often an insurer starts to fail when its aggregation goes across lines of business,” he says.

A reason why that is a challenge for them, is storing data on a client using different variations of a client’s name. That means a search for the same business throws up mismatches. Cytora automatically resolves this issue, de Castro said, by giving companies unique identifiers. “Any risk associated with a client is converted to the same ID so underwriters can identify their exposure across lines of business,” he adds.

“We have an opportunity now to use purpose-built AI solutions and really focus on specific issues and problems. If we can do that then the efficiencies we’ll see in workflows are going to be amazing” Josh Lurie, Relativity6

Understanding what a client’s business does is a “hair on fire” issue for underwriters, Lurie stresses. “Underwriters want to push a lot of volume through pre-fill submission tools but without those codes, the process can stop. Anecdotally, in the small business and micro-business world, underwriters are getting their facts wrong 50% of the time.”

Relativity6 is delivering 70% to 90% accuracy in classifying businesses depending on the segments it is analysing, he says.

And with the confidence score it provides, underwriters can “threshold and waterfall” their exposure to an insured’s risk.

Lurie concludes: “The key question an underwriter should be asking is, not how to get 100% accuracy, which isn’t realistic, but can a solution like Reativity6, delivered through a platform like Cytora’s, deliver enough of a delta to make a difference to their bottom line. We believe it can.”