AEGIS London Automates Endorsement Processing with GenAI

AEGIS London uses GenAI-driven process automation from Fisent to streamline endorsements, boost accuracy, and free underwriters to focus on higher-value risk decisions.

(Image credit: Oli Lynch/Pixabay.)

AEGIS London, which operates Lloyd’s Syndicate 1225, has implemented a GenAI-enabled automation solution called Fisent BizAI from Fisent Technologies (Toronto) to modernize its policy endorsement process. The initiative, which sits within a broader underwriting transformation program, is designed to reduce manual effort, increase accuracy, and improve workflow transparency without placing underwriting decision-making in the hands of AI.

“We started off our search for an AI provider to help us with our business process automation,” says Roger Misra, Change Leader, AEGIS London. “We wanted smarter contextual understanding of incoming, unstructured data to move our business processing to the next level of efficiency.”

Roger Misra, Change Leader, AEGIS London.

At the time of the search, large language models were becoming more mature and more accessible. “With the emergence of GenAI and ChatGPT becoming mainstream, we saw an opportunity for a more sophisticated kind of business process automation,” Misra adds.

From Slip Ingestion to Endorsements

AEGIS London initially explored use cases involving document and “slip” ingestion—the point at which risk placement information is received in the London Market. However, in working with Fisent, the company shifted its focus to a more immediate, high-value problem: endorsements.

Endorsements, which represent midterm changes to a policy, were consuming a disproportionate amount of underwriter time. Requests arrived in high volumes, in many formats and languages, often requiring repetitive analysis and manual data entry.

“We didn’t come to the table with this endorsement use case,” Misra explains. “But as we worked with Fisent, we were able to maneuver into the most value-driven use case. All of our underwriters previously had to deal with multiple midterm adjustments that were low value but high in terms of time taken to process.”

The objective was not to replace underwriting judgment, but to remove the administrative burden around it. “We are not using AI to assess our risks and we are not using AI to make underwriting decisions,” Misra says. “We’re using it to make sure the underwriter has all the information they need, in the best structured way, at the time they need it.”

Integration With Pega

Fisent’s BizAI solution is deeply integrated with AEGIS London’s Pega-based underwriting workbench, a platform the company has used for roughly eight years to orchestrate workflows.

“Fisent is very aligned to Pega [Cambridge, Mass.]. They’re a Pega partner, they understand the architecture and there is deep integration between their product and Pega,” Misra notes. “We’ve gone to them with many of our design and development problems, and they’ve suggested how we should build elements of Pega specifically relating to the Fisent use cases.”

That integration enabled the organization to automate the categorization and routing of endorsements, extracting key data elements and applying business rules that once sat exclusively in underwriters’ heads.

Rapid Maturation and Measurable Results

The endorsement automation began with a small number of rules and evolved rapidly. “We started with six rules,” Misra says of the implementation, “and in reality, once we understood the nuance of what the business was doing, we realized it wasn’t five business rules—it was closer to 100.”

Through iterative feedback with the operations team and a human-in-the-loop validation process, the system’s accuracy improved from approximately 70 percent to 98 percent in just a few cycles. Average processing time dropped from five minutes per endorsement to about two minutes, with the expectation that upcoming API integration will reduce that to 15–20 seconds per transaction, automating the majority of endorsements.

The productivity gains are meaningful. “We have removed all endorsement processing decision-making from underwriters now,” Misra says. “Underwriters never see a change to their policies unless they need to make a specific decision. We estimate that saves our underwriters about 10 percent of their day.”

Uncovering Hidden Risk

One of the most significant and unexpected outcomes of the project was greater visibility into operational risk. The system revealed that roughly 25 percent of endorsements had previously gone unprocessed, sitting in inboxes without formal capture or routing.

“That created a lack of transparency and a substantial risk to business continuity,” the case study notes. By automating intake, categorization, and routing, AEGIS London reduced both exposure and audit risk while gaining defensible, explainable processing outcomes.

Cultural Shift and Future Planning

Early adoption was driven from the top, but buy-in has since broadened to the underwriting leadership. “There was initially a push from a technology and strategy perspective,” Misra says. “Now, more and more ideas are coming from underwriters. We have two senior heads of underwriting sponsoring new use cases.”

AEGIS London is now looking at three to four new use cases per year and is beginning to think in terms of using Fisent BizAI for dozens of automation opportunities across the business.

“We want to automate all of the administrative activities that happen before an underwriter needs to touch a risk,” Misra says. “That’s the ultimate goal of where we want to be.”

ITC Briefing: Fisent Targets Automation Gaps with AI That Understands Content

Anthony R. O’Donnell // Anthony O'Donnell is Executive Editor of Insurance Innovation Reporter. For nearly two decades, he has been an observer and commentator on the use of information technology in the insurance industry, following industry trends and writing about the use of IT across all sectors of the insurance industry. He can be reached at AnthODonnell@IIReporter.com or (503) 936-2803.

Leave a Comment

(required)