How my own underwriting experience has led me to a career developing intelligent automation solutions for insurers
By Tim Crowe – Director of Insurance Solutions at Eigen Technologies
Prior to joining Eigen as the Director of Insurance Solutions, I was a Senior Vice President (SVP) at a Property & Casualty (P&C) insurance company. During that and prior stints, I gained invaluable experience advancing my career in insurance. I also realized the industry is held back by its hesitance to embrace new technology and improve workflow efficiencies. But let’s start the conversation with what I miss about Insurance.
Insurance Has Its Benefits
When you are an SVP of Underwriting at a P&C company, especially when your job involves writing specialty lines of business, you can switch companies from time to time. But what you almost never do is leave the entire industry. Why? It takes years to build relationships, so the right people send you the right kind of business in order to establish a sustainably profitable book. Once you’ve got that going, the results tend to be good pay and a great work-life balance, especially when compared to your peers in some other industries.
When it’s a hard market and rates are rising, you have more of a reason to ride that wave until the next bonus, or the next stock grant. With a business built on relationships, you’ll dine in nice restaurants and travel across the world, meeting brokers and policyholders involved in genuinely important operations often impacting global commerce. As an underwriter at a senior level, you have the authority to make decisions that without exaggeration serve as the foundation of any insurance carrier. Make the wrong ones, and your CEO will be on an investor call speaking of “balance sheet impacts”. But make the correct ones, and profits often follow, allowing the overall enterprise to flourish.
So Why Did I Leave?
The actual underwriting of risk was always interesting. But everything else surrounding that work was frustrating and antiquated, and in the many years at various stops in insurance I never saw improvement in any significant way. In fact, every year would involve some new requirement, an additional step or check that would test the bandwidth of an underwriter already under pressure to produce more premium with less.
And while the requirements of the job would increase each year, stubbornly each risk would involve essentially the same fundamental process; opening an email containing multiple documents and parsing those attachments for the material info I needed in order to quote the business. But before I could quote, I would take the info I extracted manually and rekey it into some pricing rater. Or I would rekey it into an aggregation system. Or perhaps I would rekey that aggregation information into an email that I would then send to a third-party (who would, presumably, then rekey that info into their own systems).
Sometimes I would rekey info into an email for an authority sign-off on a particular risk, or maybe to purchase facultative reinsurance, or maybe to request authority sign-off to purchase that facultative reinsurance (...this is the stuff I don’t miss). And at some companies, even the quote creation step was a rekeying exercise itself, involving opening a Word template and manually filling in various required fields, converting to a PDF, attaching to an email, then dragging that email into a document management system followed by the manual labeling of the file. Honestly, I am not making this up! And as the years went by, and the ability to, say, order food and/or taxis became painless and almost instant in the “real world”, the delayed nature of these insurance processes became increasingly unacceptable to me.
How Can Things Still Be Like This?
The insurance industry generally recognizes (lack of) efficiency as a real concern across the value chain. Unfortunately, large amounts of capital have been spent on “solutions” (such as shiny pricing tools and the like) that don’t address the root cause of inefficiency, namely the need for a human or group of humans to review multiple documents plus associated rekeying before downstream processes can commence. This is the bottleneck. I repeat, this is the bottleneck.
Until this bottleneck is addressed, the speed of any quoting system, for example, cannot exceed the speed at which a person can sort through documents and manually enter requisite information into whatever systems the quote process relies upon. Some of this process can be outsourced or handed off to assistants, but no matter how much money or people you throw at it there will still be latency, and that is before considering the potential issues associated with accuracy and quality.
Another reason in my opinion is the quality of vendors focusing on insurance can be a bit lacking, particularly in ‘specialty lines’. While there are quite a few genuinely good quality companies, my experience was those were mixed with vendors who would promise big things that they would ultimately fail to deliver. And when a solution to a problem involves things like cutting-edge machine learning, how would an insurer determine who is for real, and who is overpromising?
What can be done about it?
I was fortunate several years ago to have had a professor at graduate school who convinced me to learn how to code (badly as it turned out, but that’s for a future blog). That led to the studying of machine learning, an attempt at a direct-to-consumer startup to address the power imbalance in insurance transactions, and an overall growing belief that those machine learning techniques, including natural language understanding (NLU) and computer vision, utilized in partnership with human experts, could improve the transactional efficiency of insurance (i.e., the thing that paid the bills for most of my working career).
During my time as an underwriter at various stops, I would contemplate how to speed up the initial review of submissions, and how a system could provide an underwriter with, say, only those risks potentially in appetite, while auto declining the rest. That could save quite a bit of time, I would think to myself. Maybe a system would include an easy-to-use interface that was pre-populated with information powered by automated extractions. A bit like how, when you scan a check at an ATM, you get a pre-populated verification screen where the deposit process is contingent on you agreeing to the extracted values.
Perhaps quoting could be sped up along with improved accuracy levels, even though actual underwriter effort levels could stay the same or even be reduced. A lovely theory, and one I still believe. Ultimately, I determined that instituting this type of change would only be possible by leaving the industry and joining a company which had the technical capabilities that most aligned with my overall vision.
At Eigen, we deliver on our promise to enable any user to automate the extraction of information from multiple documents to save time, money and human effort as well as reduce errors across multiple functional areas and use cases, including underwriting.
- World Economic forum 2020
- Gartner Cool Vendor 2020
- AI 100 2021
- Lazard T100
- FT Intelligent Business 2019
- FT Intelligent Business 2020
- CogX Awards 2019
- CogX Awards 2021
- CogX Awards Best AI Product in Insurance
- Forbes The AI 50 2022
- Ai BreakThrough Award 2022