Assured Automates Information Collection For Auto Insurance Claims
October 1, 2020
Getting into a car accident is frightening. Having to explain what happened to your auto insurance company can be even more stressful. Claim specialists spend exorbitant amounts of time and effort to get all the necessary information from the claimant. The more time spent recording key information needed to process a claim, the longer it takes for the claimant to get reimbursed. Justin Lewis-Weber and Theo Patt saw the locked value behind this archaic process, creating Assured to fix it. Assured is an insurtech startup creating claims automation technology to streamline processing insurance claims. The San Francisco-based company has raised money from Global Founders Capital, Neo, and Henry Kravis.
Neo CEO Ali Partovi says, “I love betting on founders like Justin and Theo who dare to reimagine the status quo. Assured is on track to enable a breakthrough that has eluded the insurance world for years, and the team’s industry veterans validate its viability. Assured’s digital claims processing promises a better experience for consumers and a paradigm shift for the industry.”
Frederick Daso: What led you to discover the $70B problem of processing insurance claims in the U.S.?
Justin Lewis-Weber: As an entrepreneur, I’ve always sought to disrupt industries that I see as core to how society functions, but have been traditionally overlooked. Insurance is an excellent example of this—car insurance is one of the few private products legally required in the U.S., yet Silicon Valley has paid it relatively little attention.
Before I started Assured, I discussed with a friend in the insurance space who mentioned—too casually—that a full 10% of Property and Casualty premium goes to claims processing, excluding the actual claim payout. Across the U.S., with $680 billion of premium written annually, this adds up to nearly $70 billion a year, basically answering:
Is this fraud?
Who is at fault?
How much should we pay?
I found this fact incredibly surprising—and developed a set of theses around how to reduce this cost by at least 10x. These theses became what Assured is today.
Daso: In our previous discussion, you mentioned that previously tried solutions using artificial intelligence, machine learning, or natural language programming have failed. Why did they fail, and what did you specifically learn from those failures that shaped your views on a potential solution?
Lewis-Weber: One of the first things I did when diving into the insurance claims space spent a considerable amount of time personally in claims centers, listening to phone specialists intake and process auto claims.
I immediately saw that the underlying claims data being used to process the claim was being ingested in a very manual way—over about an hour of phone calls with the claimant, spaced over many days. The phone agent paraphrased what the claimant said into basically a giant text field called “Claims Notes,” and that was it!
It was no wonder to me why previous attempts at solving the problem had failed: they all centered around leveraging the narrative-style “claims notes” and Natural Language Processing (NLP) algorithms to produce a human-like understanding of the claim.
Unfortunately, this isn’t possible with even state of the art NLP. The core insight of Assured is that to achieve this end game of automated claims processing, the claims data—from the start—has to be gathered in a structured and ultimately machine-readable way.
We do this at First Notice of Loss (FNOL)—i.e., the first interaction where the claimant reports the loss. We replace these numerous phone calls with call center agents with a beautiful and intuitive web app that ingests the claim information in a highly structured way.
In this way, we directly enable A.I. and Machine Learning based claims processing to succeed while also providing a dramatically better customer experience.
Daso: Why is it so expensive to process an auto insurance claim? What is the main cost driver?
Lewis-Weber: To put it concisely, claims processing is expensive because it’s highly manual. More than a quarter-million Americans are employed as claims adjusters—that’s nearly 1 out of every thousand Americans! These people are highly trained, expensive to hire, and can only process so many claims a day. By automating the FNOL process, we save these adjusters time (enabling them to process more requests per day), and Assured’s highly-structured data model paves the way for straight-through claims processing.
Daso: How are premiums structured to account for the cost inefficiencies in processing these claims?
Lewis-Weber: Frankly, they’re just higher! The basic insurance model is that we all pay into the pot and those of us that need it to take money out of that pot (claim payout). However, in actuality, aggregate premiums will always outweigh aggregate payouts because insurance companies take money to operate and need to make a profit. Decreasing the cost of processing claims is money that insurers can use to reduce premiums, allowing them to undercut their competitors and put money back into consumers’ pockets.
Daso: You have built Assured to focus first on the problem of ingesting the data. What are the specific facets of the relevant data you’ve focused on collecting? How do you then restructure this data into a machine-readable way that not only allows for easier processing but is intuitive to a user that relies on it to process a claim?
At Assured, we aim to give both human and machine adjusters a “situational awareness” of the claim. That is, everything you’d come to know if you were told the story verbally, but in a structured way and without requiring the claimant to enter lots of text or talk over the phone.
This is complementary to in-car telematics (like dashcams) because no matter how many cameras you have, you still need to get “the story” from the driver.
We do this by focusing on making the questions streamlined and simple for users to answer. We’re incredibly proud of the fact that there are precisely zero text fields in our flow. Instead, we ask multiple-choice questions that are highly informed by everything we know about the claim and claimant. We integrate more than 50 external data sources (things like weather conditions and road geometry) and utilize technologies like Computer Vision and Optical Character Recognition.
For a simple consumer experience, there’s a lot of complexity beneath the hood. Dependent on users’ answers, there are more than 8.55 million different flows they might experience—all designed to improve user experience and the utility of the gathered data.
Daso: How did you determine that the First Notice of Loss (FNOL) could be an automated step? What insights did you glean from studying the traditional process of multiple calls with customer support?
Lewis-Weber: What struck me most about listening to traditional FNOL calls is both how different the calls were between agents, and how similar the underlying gathered data was. The question set had to be standardized, but also made much more specific. The most common question asked was, “What Happened?” Of course, this is super broad, and a regular consumer doesn’t know what the adjuster is looking for, so they give this long rambling narrative that tries to cover all their bases. This is unnecessarily stressful and makes it almost impossible to structure their answer into data fields. By utilizing specific, yet highly agile and dynamic question sets, I knew we could do better.
Daso: You have an aerospace background like me. How do you use your aerospace background to break down the problems you face at Assured?
Lewis-Weber: That’s exactly right, Fred! I view an insurance claim as this deceptively simple problem with huge amounts of hidden complexity. It’s simple because, in some sense, it’s objective. There is a set of policies and business rules, and it’s our job to logically go through and match the situation to the set of rules.
But there’s a massive amount of complexity in accurately understanding the situation well enough to match it to the set of rules. And to me, that’s a super fun and difficult engineering challenge. We solve the problem by making the situation more objective and structured, and therefore easier to understand. There are tons of tradeoffs: the more questions we ask, the better we can understand the situation, but the more time it takes the customer to file their claim. We have creative solutions to this problem, such as using ML mid-flow to understand fraudulent behavior and ask more questions but giving trustworthy users a quicker flow.
This is a lot like designing an aircraft or spacecraft. All of these compromises, and having a great outcome at the end of the day, comes down to intelligent and intentional engineering tradeoffs.
Daso: You mentioned your cofounder Theo is incredible in his own right, having started a company previously like you. What lessons did you both learn from your previous companies that have shaped how you two are building Assured for the next ten years?
Lewis-Weber: Theo is an incredible engineer and entrepreneur. Like me, he has started several successful companies before, both in the consumer and enterprise space. That gives us both unique insight into building a product for the enterprise (insurers), that still pays a considerable amount of attention and respect to the consumer (claimant) experience.
As serial founders, there’s a large part of company building that comes much more comfortable because you’ve done it before. That makes it way easier to focus on our customers and product and ultimately allows us to add more value to our customers faster than less experienced founders can.
Ultimately, having the dream and long-term vision for Assured is the easy part. However, none of that matters if you don’t deliver vast amounts of value to your end customers. And for the Assured team, customers are always at the heart of everything we do.
For the latest tech news, subscribe to my newsletter, Founder to Founder.