Can AI be trusted for premium finance planning?

Artificial intelligence has become so advanced that it can now reliably be used to process and make decisions in complex processes such as premium finance planning, according to several industry experts who spoke with InsuranceNewsNet.

“Bringing AI into the conversation in terms of modeling — looking at your entire portfolio, determining stress testing and analysis — you can, of course, use artificial intelligence, specifically machine learning, to understand default probabilities of certain aspects of that portfolio,” Franklin Manchester, global insurance strategic advisor, SAS Institute, explained. “Probabilities can be higher or lower, depending on the risk associated with the profile of the consumer to whom the money is being lent.”

AI’s prevalence in insurance is expanding, with use cases quickly advancing from simple automation of repetitive tasks to advanced reasoning such as premium financing and even claims settlement. 

However, AI’s inherent related risks still raise major concerns in an industry predicated on risk management and aversion.

Manchester noted that the waters aren’t completely clear just yet. He raised caution about the “adverse outcomes” that could result if technology is not applied appropriately, such as if an organization that offers premium financing decides to increase rates or not to lend out funds after all “because of how the AI was ultimately used.”

“Now, I’m not going to sit here and say that’s right or wrong; it is the terms of the agreement that are acknowledged by that individual when they are choosing to secure that financial vehicle to pay their premium. However, the business or the consumer does not have that insurance protection. So, ultimately, if the AI modeling is used for the premium finance company to go through that stress testing term and risk stability and those decisions are made, that is an adverse outcome,” Manchester said.

Assessing complex scenarios

Henry Zelikovsky, founder and CEO of Softlab360, explained the mechanism behind how and why technology can be used to improve scenario analysis in premium finance planning.

He said machine learning models first “examine data holdings of a portfolio over a period of time and cash distributions of a portfolio over a period of time to determine what would be a reasonable recommendation and predict what would be a better option given portfolio composition, what can be freed up to raise money for a premium.”

Then, conventional conversational AI tools such as Claude or ChatGPT can be leveraged above the structured numerical data that’s predicted through the language models to “really allow a personal conversation.”

“Someone could take advantage of the Claude model as a large language model to phrase a question: ‘What should I do? What can be my options? What could be a better option?’” Zelikovsky said.

Pavel Sukhachev, founder of Electromania, maintains that companies don’t necessarily need a complex AI tool for this purpose. His organization uses Claude for 99% of coding and finds it perfectly suitable for their needs. 

“I’m excited about this because it saves you a lot of time. Previously, for us to do some kind of deep research, we’ve had to identify a lot of sources, contact all the sources, grab the information and compare it. It’s been quite a challenging thing. Now with the help of AI, there’s no bottleneck at all,” Sukhachev said.

Zelikovsky acknowledged concerns about how reliable or accurate AI can be when performing premium financing analysis. However, as a former engineer who has spent the last 10 to 12 years in machine learning through his business, he believes leveraging AI in this way is reliable because it merely enhances an assessment process that has existed for years. 

Further, he emphasized that machine learning holds true to its name — it’s always learning, processing and analyzing historical data as well as new data and learning from that database of information.

“I think concerns are valid. I think we will use the result of what’s called an
autogenerated conclusion and the question is, was that accurate?” Zelikovsky noted. “The companies that do this for a living, like my company, would come with the tools to prove that the result is consistent, and the person who looks at the results would contribute their own opinion of what they saw as a result of such an AI to decide whether it’s an appropriate conclusion.”

Insurance-specific AI solutions

Like Zelikovsky, Manchester represents one of the many technology companies that design AI solutions specifically for the insurance and finance industry, including options to support premium financing analysis.

“The AI that we offer and the AI that is generally available can do these very complex risk calculations fast, better than humans can with one customer — a different financial vehicle, but variable annuities,” Manchester said.

For instance, he said SAS saw a 99% reduction in stress testing calculations for expected cash flows. He added that the power of their Risk Engine tool versus standard analytical tools or even Excel — which he suggested some companies are still using — is precisely this ability to effectively “go through that scenario analysis very quickly.”

Zelikovsky, meanwhile, said his company’s LLM follows a rigorous training protocol of learning and relearning using historic data from 20 to 30 years of portfolio history until consistency reaches a benchmark of 94% or higher.

Throw new data in the mix for that AI to learn from and support it with human talent evaluating the conclusions against the machine that made the prediction, he added, and you have Softlab360’s formula for success.

“That’s our concept, and we’ve been doing it for a while. We’re a software engineering company of a solutions nature; we use our tools, our techniques to provide implementation of customer use cases per customer,” Zelikovsky explained.

Overseas, there’s an even more specific solution being developed — a new AI-powered platform designed exclusively for premium financing analysis and support. The platform is aimed at providing a loan within a matter of minutes, facilitating customizable repayment options and providing a “white-label” solution for insurers to have more control over the process instead of having to give up a big part of their brand to third-party companies. 

Enrico Damioli, co-founder, explained that his new organization, Flexra, is building an API-first platform that any insurer or broker can easily integrate to give them more control.

He said its development was partially a response to a “very concentrated” market in the United Kingdom, where there are only “a few large players” in a fairly “competition-free” market with clunky systems and seemingly little motivation to improve. It’s in the prototype phase as of this writing but is expected to fully launch within the next 10 to 12 months.

“The idea is to have a platform that can be used by more old-school types of insurers, managing general agents or brokers. So, an external platform that they can use to monitor all the premium financing that they have, create new premium financing, get in touch with the clients and then have a platform to have all of this under control,” Damioli said.

However, Zelikovsky underscored that regardless of what solution finance providers opt for, what matters most is evolving the conversation about AI and how it is used.

“My vision, and suggestion to the industry, is that it is important to evolve the conversation,” he said. “It’s important to make sure that people try them out. There is no going back. AI is a fundamental here.” 

Rayne Morgan is a journalist, copywriter, and editor with over 10 years' combined experience in digital content and print media. You can reach her at [email protected].