The Center for Data Innovation spoke with Paul Monasterio, co-founder and CEO of Kalepa, a commercial underwriting software company based in New York City. Monasterio explained how underwriters can use data to better assess prospective insurance applicants and discussed the impact of AI on the insurance industry.
Morgan Stevens: What was your motivation for starting Kalepa?
Paul Monasterio: It’s an interesting journey. My co-founder and I come from two different worlds and neither of them are in the insurance space. I have a science, data, and technology background while he comes from a military intelligence background in the Israeli Defense Forces. We met at a company called Applied Predictive Technologies, which was a part of the early generation of AI companies. There we worked closely with many insurance companies across the entire value chain, such as distribution, underwriting, and claims, and viewed the entire process through a strong data and technology lens. We also gained an appreciation for the mission of the insurance industry and how people and businesses can rely on insurance companies to take care of them when things go sideways. We saw some of the problems within the industry and learned how data and technology could help solve them. When we decided to start Kalepa, we quickly realized that underwriting was ripe for taking advantage of many of the new technologies that have emerged over the past few years. As a result, we created Copilot, an underwriting platform that can help every underwriter do their work better and faster, with the ability to prioritize their work, look deeper into risks, and understand exposures and controls in ways they have not been able to do before.
Stevens: Can you explain the technology behind Copilot? How does it empower underwriters in their risk evaluations?
Monasterio: The job of an underwriter has many facets. It’s a tough job to do. An underwriter needs to take submissions and a variety of documents with valuable information and understand exactly what a prospective insured is looking for, the exposures, and the controls, or factors that can mitigate potential risks. They then need to compare those factors with their appetite and portfolio goals. Insurers and underwriters are ultimately trying to move the needle on the growth of the book but also need to ensure that they do profitably. There’s a way to accelerate growth by saying yes to everything, but that typically leads to adverse selection and a very poorly performing portfolio. Alternatively, insurers and underwriters can say no to everything and guarantee that they won’t have a single claim, but they likely won’t stay in business for very long.
Copilot helps the underwriter at each of those steps. As submissions come in, it uses a variety of techniques such as optical character recognition, natural language processing, and large language models, many of which are growing in their strength and capabilities, to turn structured or partially structured data, such as application documents, loss runs, and supplemental information, into something more useful. It’s essentially a software tool that can handle ancillary tasks to support human underwriters.
It also helps underwriters prioritize which submissions they should work on. Not all submissions are created equal. The one nonrenewable resource underwriters have is time and Copilot helps them use it more efficiently. In addition to initial information in submissions, Copilot can take information from the Internet, news articles, and government databases and help underwriters better understand applicants’ risks. It recommends which businesses are more likely to bind profitably and fulfill book requirements and can identify exposures and controls. An underwriter might receive submissions from two businesses that look similar but are actually very different. If they use a broad stroke to make a decision, they might treat both businesses unfairly. They can believe that one business is actually riskier than it is and give them too high of a rate or deny a quote. Conversely, they could believe that a business is less risky than it is and have a claim experience that is worse than anticipated and might cause their rates to rise across the board. For example, a restaurant in Washington, D.C. may apply for a policy. Using Copilot, an underwriter may discover that the restaurant has a mechanical bull in the back from Yelp or Google reviews. The restaurant would have a significantly different exposure than a family restaurant that simply serves food.
Stevens: What differentiates Kalepa from other organizations in the insurtech space?
Monasterio: The insurtech space is very broad. There are many organizations that are actually insurance companies that are taking new approaches to writing insurance and serving customers. We’re a different flavor. We are a software company that empowers large insurance carriers, small insurance carriers, managing general agents (MGAs), new entrants in the industry, and more to address problems with underwriting and enable underwriters to be more efficient. Many software companies have tried to tackle the issue of underwriting over the years but our key differentiator from them is the ability of our product, Copilot, to bring all the necessary features together to deliver on the key goals of growth and profitability. In underwriting there are several significant technical problems, such as entity resolution and making sense of complex documents or images, that need to be resolved correctly in order to help an underwriter make the right decision quickly and efficiently. Copilot cracks a lot of those problems in one seamless experience–and one that automatically improves and adapts to the underwriter.
Stevens: How can the insurtech industry use AI to prepare for potential losses stemming from climate change or related events?
Monasterio: There are a number of ways that insurance carriers can take advantage of new technologies to prepare for more frequent natural disasters. In general, carriers can use a greater availability of data, at a subtle level that has never been seen before, to better assess risks. Companies, especially property insurance companies, can analyze high quality satellite images to get very granular reads on foliage coverage, urban forest separation, and more. Taking that information and turning it into signals around risk scores is a massive opportunity that can inform underwriters about properties with higher risks of experiencing a natural disaster. By using some of the latest techniques to analyze imagery data, underwriters can render a much more accurate and granular assessment and give individuals and businesses that have taken the right precautions preferential rates. The world cannot avoid natural disasters. By definition, they will happen, and they will happen unpredictably. But insurers can take advantage of technology to better understand how to build a portfolio and underwrite in a manner that takes into account these events and individuals’ and businesses’ preparations for them.
Stevens: How do you see AI transforming the insurance industry over the next decade?
Monasterio: At Kalepa, we’re big proponents of the idea that AI is about augmentation instead of automation. Some people are worried that AI will replace human workers and we don’t subscribe to that notion. There are some things that machines do very well and some things that humans do very well, and those are not the same things. Machines can accomplish some tasks, such as looking at 10,000 data points in a matter of seconds, better than humans, but they have awful judgment. Instead, insurance companies that embrace augmentation and utilize both human workers and machines for underwriting will reap rewards. We’ve seen some organizations say that they don’t need an underwriter and can instead automate the process with AI. They typically end up with an extraordinarily bad book of business filled with unsurprisingly silly mistakes. On the other hand, companies claiming that AI is all buzz are also incorrect. There is a lot of hype around the technology, but it has significant potential. It is critical to separate the wheat from the chaff. We expect that companies that take advantage of AI and use technology to help their human workers accomplish more tasks faster and more effectively will deliver the most return to shareholders and better advance their corporate objectives.