Recently Risk & Insurance editor-in-chief Dan Reynolds caught up with Chaz Perera, CEO and Co-Founder of Roots Automation. What follows is a transcript of that discussion, edited for length and clarity.

R&I: If you just don’t mind talking about how Roots Automation the company came to be and what the motivation was for it?

CP: Sure, I don’t mind talking about how Roots Automation came to be! My co-founder John Cottongim and I both had long careers in the insurance industry, and we saw firsthand the challenges that insurance professionals faced in their day-to-day work. At AIG, for example, we struggled to keep talent because people were bogged down by tedious tasks and weren’t excited to be spending so much time focused on menial and repetitive tasks like data extraction. One of the things we realized was that insurance processes rely heavily on unstructured data, like emails, submission applications, and ACORD forms. This made it incredibly difficult for underwriters and claims adjusters to do their jobs efficiently and accurately, and it also kept them from engaging with customers.

We knew that if we could solve the unstructured data problem, we could help insurance companies transform and become more customer-centric. We could free up people’s time to get on the phone, engage with brokers, engage with policyholders, and engage with claimants. So, we started Roots Automation in 2018 with the goal of using AI to automate these manual processes and free up insurance professionals to do more meaningful work.

R&I: How does Roots Automation solve the unstructured data problem?

CP: The problem with unstructured data in insurance is that it’s incredibly difficult for a human to accurately read through all of it and get the key data points they need to assess risk and make decisions. This can lead to things like claims leakage, missed deadlines, and undercharging for premiums.

At Roots Automation, we use AI to solve this problem. We’ve built a mosaic of traditional machine learning models and fine-tuned large language models that work together to split, classify, summarize, and extract information from unstructured data sources. Once we ‘freed the data’ we solved the next challenge of how to use the data to automate processes.

Our insurance-specific AI software, called InsurGPT, is trained on millions of proprietary insurance documents and is constantly learning and improving. This allows us to automate manual processes and provide insurance professionals with the accurate and timely information they need to do their jobs effectively.

R&I: What makes Roots Automation’s approach to unstructured data unique?

CP: I think what makes our approach to unstructured data unique is that we’ve built a platform that is not just about reading a document. It goes beyond old-school optical character recognition (OCR) and handing you back out-of-context and lower-quality results. We’ve developed digital coworkers who can actually understand the context of the data and use that understanding to automate complex tasks. We don’t just host and run models, we also manage workflows, transact data, interact with screens, summarize items, and make suggestions.

Our deep understanding of insurance problems and our experience across the full tech stack allows us to create solutions that deliver real value to our customers. We spend a lot of money on research and development to ensure that our models are constantly learning and improving. For example, we train our models on millions of insurance documents and constantly fine-tune them to ensure the highest levels of accuracy. Our goal is to provide our customers with a complete solution to automate their end-to-end processes and help them transform their businesses.

R&I: Where are Roots automation customers seeing improvements?

CP: We’ve seen that insurance carriers can quote more opportunities because we can accelerate the time needed to produce a quote. It is not uncommon for an insurance carrier to quote somewhere between 20-25% of the opportunities they see. Many submissions are declined because they don’t fit within the carrier’s underwriting guidelines, but a lot of other requests just don’t have the time to react to and get processed. We can also do it faster and more accurately than a human, which helps to reduce the premium miscalculations that are occurring that lead to premium leakage.

On the claims side, we allow claims departments to increase their capacity. It is pretty common that on the first day our AI digital coworkers ‘show up for work’ at an insurance carrier, they see an almost 90% increase in capacity. Meaning if they have 10 people doing the job, nine of them can now focus on something else. Beyond that capacity gain, we’re seeing improvements in claims leakage because our insurance AI software is identifying time-sensitive information that when not acted on leads to overpayment of claims.

For example, in workers’ compensation where you might get a request for authorization from a doctor… “Will you pay me to perform this procedure?” You typically have five days to respond. If you don’t respond in five days, you theoretically owe them that money, whether you thought the treatment was warranted or not. And it’s not uncommon for adjusters to miss that about 20% of the time.

Another common example are time limited demand letters. Essentially, the lawyer sends something to the adjuster saying, you’ve got to respond to me in 30 days, and if you don’t, I’m taking you to court. They’ll bury that demand request in a single file with a couple hundred pages of medical records, PT updates, and a whole bunch of other things. Humans will miss these at least 25% of the time. And if you imagine you could have settled a claim for, say, $50,000, but now that it’s going to court, your litigation costs may be $50,000, never mind the indemnity and everything else. So, it’s those types of areas that provide concrete value for insurance companies.

R&I: Are you getting feedback that what you’re doing is improving the work environment for adjusters and underwriters?

CP: The most common thing I hear is that adjusters are now able to leave work on time because our digital coworkers are taking care of a lot of the tedious tasks that used to keep them in the office late. For example, one adjuster told me, “I actually get to leave at 5:00 or 5:30 every day. It used to be that I would hang around, or I’d be stuck at home working in front of my desk till 6:30, 7:00 to get through everything so I was prepared for the next day.”

Another thing I hear a lot is that insurance people feel less stressed on Sundays because they know that when they show up on Monday, their work will be organized and ready for them to go. They’re not worrying about having to sift through hundreds of pages of documents to find the key information they need. This is really important because insurance is a relationship business. It’s a people business and insurance professionals need to be able to focus on building relationships with their customers, not getting bogged down by paperwork. They don’t have to work late or on the weekends and can make it to the soccer tournament.

R&I: Given your knowledge of the market, where are you seeing insurers make technology investments?

CP: I am seeing insurance companies make heavy investments in self-service. We are seeing some self-service applications and support appear for commercial buyers in addition to the more common self-service offerings available to policyholders on the consumer side. This is important because the Millennial demographic really wants to be able to do their research on their own before speaking with an insurance professional.

I am also seeing that insurance companies are finally making investments in AI to solve the unstructured data problem. Our digital coworkers can read any type of document in any shape and size, including emails and attachments, and extract the data points that insurance professionals need to make decisions. This is leading to improvements in efficiency, accuracy, and customer satisfaction.

R&I: One of insurance’s biggest hurdles is that people don’t understand how insurance works. Right?

CP: I agree with you, it does seem like a big hurdle in the insurance industry is that people don’t understand how it works. I used to tell my team at AIG that part of the problem is that there’s this vacuum that occurs between when a policy is bound and when a claim occurs. During this time, customers aren’t really interacting with their insurance provider, which creates uncertainty. This makes it harder for them to see the value in their policy. Unlike Apple, where you interact with their product thousands of times a week, theoretically, insurance is lucky if they get an interaction with their customer four or five times a year. Often, those interactions are emotionally charged after something bad happens, like a car accident.

A lot of people think that insurance companies are competing on price, but I’ve found it’s more about service and relationships. As a business owner, I’m happy to pay more if I know I’m getting great service. I think if insurance companies can focus on providing better service and explaining how they’re working to improve the customer experience, people will think more favorably about insurers.

R&I: Is there anything about the company, Chaz, that you think is really important that we haven’t covered yet?

CP: I think one of the things that’s really unique about Roots Automation is that we have a great reverence for the insurance industry. My co-founder and I lived in the industry, we grew up in the industry, and we learned everything about the business from within the industry. We didn’t come at this from the perspective of just wanting to solve a technology problem. We saw a problem that was preventing insurance professionals from doing their best work and providing excellent customer service, so we set out to solve it.

That’s why we’re so focused on building solutions tailored to the insurance industry’s specific needs and that can help insurance companies transform their businesses. We’re not just a technology company; we’re a technology company filled with insurance people who understand the great care that is needed to make insurance the insurance product successful. &