The insurance industry is looking to adopt artificial intelligence applications for a variety of business functions due to its access to large volume of customer data. According to our AI Opportunity Landscape research in insurance, approximately 46% of AI vendors in insurance offer solutions for claims and 43% offer solutions for underwriting. While there are other areas of insurance that could benefit from AI, it’s clear that these two core insurance tenets are where the traction is with AI in the industry.
We researched the space to better understand where AI comes into play in automating claims and the underwriting process. This report is a snapshot of our full AI Opportunity Landscape in insurance, but it aims to answer the following questions:
- What types of AI applications are currently being used for claims processing and underwriting?
- What are the tangible results of AI-driven for insurers seeking to automate claims and underwriting?
- Are there any common trends among these innovation efforts? How could these trends affect the future of the claims and underwriting processes for insurers?
This article intends to provide business leaders in the insurance space with an idea of what they can currently expect from AI in their industry. We hope that this article allows business leaders in insurance to garner insights they can confidently relay to their executive teams so they can make informed decisions when thinking about AI adoption.
At the very least, this article intends to act as a method of reducing the time business leaders in insurance spend researching AI companies with whom they may (or may not) be interested in working.
Claims Processing
Lemonade
Lemonade is not a B2B AI vendor, but instead an insurance company themselves. They claim they can process claims faster and provide customers with fast payouts using AI, including a chatbot.
Lemonade claims users can tap a button on the Lemonade mobile phone app in order to begin the claims process. The user is brought to a messaging application where they can explain their situation to a chatbot named Jim. Then, it seems the application uses AI to match the claim description to similar descriptions stored in its database, searching for any identical claims in order to determine if the claim is fraudulent.
If the system determines the claim to be legitimate, the company says it will automatically approve the claim if it isn’t too complex. This is likely done in the same manner as fraud detection, matching the complexity and value of the claim with similar claims stored in Lemonade’s database.
If the claim is deemed to be too complex, the chatbot will inform the user that a customer support representative (a human) will contact them as soon as possible. Although the company does not make this clear, we can infer that a customer’s payout can be deposited into the user’s bank account.
Although the company does not provide case studies in the way we often report them (due to the fact that they are not a B2B vendor), Lemonade does claim that a customer was paid within 3 seconds of his claim being approved.
That said, the customer had filed a claim for a coat worth around $900, and so the claim was more straightforward than claims one might find in health and auto insurance. Lemonade themselves admit that Jim can’t handle more complex claims and that such claims are elevated to “real Jim.”
Below is a short 1-minute video demonstrating how Lemonade’s claims process works:
Shai Wininger is co-founder at Lemonade. He is currently a board member at Fiverr, which he also co-founded. Previously, he was the founder and CEO of Handsmart and Trimus Inc.
Tractable
Tractable offers a software which they claim can help insurance agencies automate claims using machine vision.
Tractable claims insurance agents can upload images associated with the claim, such as those of a damaged car, and an estimate of how much they think the client should receive as a payout based on the damage in the image. Then, Tractable’s software compares the uploaded image to a backlog of images of various severities of damage and the payouts associated with them.
The software checks if the insurance agent’s payout estimate is more than the payout that other clients received for similarly damaged vehicles.
If it is, the software notifies the insurance agent, and the agent can then decrease the payout that they intend to provide the client based on the payouts that past clients received for similarly damaged vehicles. Thus, Tractable claims their software can reduce the amount of claim leakage when agents pay claims.
In the video below, Tractable’s co-founder and CCO, Adrien Cohen, demonstrates how Tractable’s software works between 2:18 and 9:45:
Our research yielded no results when we tried to find case studies for the software. We caution readers to take this into consideration when determining potential vendors.
Tractable does not list any major companies as clients, but they have raised $34.9 million and are backed by Insight Venture Partners, Ignition Partners, and Zetta Venture Partners.
Razvan Ranca is CTO at Tractable. He holds an MS in computer science with a machine learning distinction from the University of Cambridge.
Elafris
Elafris also offers a chatbot-based software to insurance agencies so that they can help customers make monthly insurance payments and check the status of their claims.
Elafris claims customers can open a chat window with an Elafris chatbot on their cell phone, starting the conversation with something like “view existing claims.” The chatbot will then return the customer’s existing claims, and the customer can select one of those claims in order to be sent information about it, such as if the claim was approved, the payout amount, and the payout check has been mailed out to the customer.
Elafris claims the customer can also begin a new claim within the same chat window, relaying the details of the situation, such as what was broken and when. After answering all of the bot’s follow-up questions, the customer will have submitted their claim in full to the insurance company.
Then, the customer would then wait to be contacted by an insurance representative. The chatbot Elafris builds does not seem to be able to approve claims the way Lemonade’s might, but Elafris claims their chatbot works through Amazon Alexa in addition to traditional messaging applications.
Below is a short 2-minute video, found on Elafris’ website, demonstrating how the chatbot works to allow clients to file a claim:
Elafris was founded in 2016 and seems to be in its very early phases of funding, but we could not find any exact number as to how much they have raised so far nor from whom they raised that money. As such, their service does not seem to be on the market yet.
Elafris does not make available any case studies reporting success with their software, but they do list clients such as Jupiter Insurance, Rightsure, and Majesco under their list of “Clients and Partner.” It is unclear what exactly the details of “partner” status entail for Elafris.
We were also unable to find any C-level executives with AI experience on the company’s team. Time will tell if the company can secure funding from Silicon Valley firms, which will likely require their software to have a robust foundation in machine learning or natural language processing (NLP). Although the company claims to be doing AI, their chatbot could very well work on if-then scenarios, which although might be AI, is neither machine learning nor NLP.
Underwriting
Cape Analytics
Cape Analytics offers a service which they claim can help property insurers underwrite more accurately and more cost-effectively using satellite-based machine vision.
Cape Analytics claims insurers can provide the company with an address. Then, Cape Analytics’ software uses machine vision to analyze satellite images of the property at the address.
The system then seems to provide the insurer with a list of aspects of the property that may be of interest to them, such as details about roofing; details about any existing pool, including how it’s enclosed and if it has a diving board; trees on the property that may be prone to falling or catching on fire; or whether or not the property has a trampoline on it.
This could provide the insurer with an accurate inspection of the property without requiring the hiring of an actual human inspector that shows up at the property.
Likely due to the nature of the software, we were unable to find any demonstration videos for how the software works. The company’s website only provides a marketing video.
Cape Analytics does not provide any robust case studies on its website, but some case studies touting the company’s success are provided by Oxbow Partners. These case studies are short and lack detail, and so we suggest readers be cautious about accepting their claims.
For example, Cape Analytics claims to have helped an unnamed regional US insurer reduced inspection spend by 50 percent while using Cape Analytics. In other words, the company was able to partially forgo hiring inspectors to show up at properties and verify the existence of aspects of the property that would affect the underwriting decision.
Cape Analytics does not list any large or marquee companies as clients, but they have raised $31 million from several Silicon Valley investors.
Suat Gedikli is CTO and co-founder of Cape Analytics. He earned his Ph.D., in computer science, image processing, and probability state estimation from the Technical University of Munich. Previously, he worked as a research engineer at Willow Garage, which was the progenitor of several companies, such as Industrial Perception and hiDOF inc. The latter was acquired by Google.
Daisy Intelligence
Daisy Intelligence offers a software which they claim can help insurance agencies automate the underwriting process with machine learning by providing price suggestions for different customers based on their individual risk factors.
Daisy claims users first upload at least two years of “operations data” into their software. We can infer this data includes historical customer data, such as their ages, blood pressures, locations, etc. Although Daisy Intelligence is not explicit in explaining what the software does with this data, we can infer based on similar software that the system finds patterns among the insurance agency’s existing customers that inform their risk.
Then, the software provides pricing recommendations for individual customers based on those patterns. These recommendations are, according to the company, updated weekly. Daisy Intelligence says that their staff can assist in executing those recommendations, which may mean that the system requires a robust integration process and some kind of regular contact with the vendor in order to learn or maintain it.
The company does not provide any demonstration videos for its software.
Similar to many of the companies in this report, Daisy Intelligence does not offer case studies for its underwriting software, but it does provide details regarding success for their retail solution. Again, not unlike other companies in this report, they do not list any clients on their website for their underwriting software. They’ve raised an undisclosed amount of money from JSM Capital and Manjis Holdings.
Sina Meraji is Director of Machine Learning and Software Development at Daisy Intelligence. He holds a PhD in computer science from McGill University.
Planck Re
Planck Re offers a software which they claim can help insurance agencies automate the underwriting process with machine vision.
Planck Re claims insurance agents can type in the name of a business and its physical address in their system. Then, Planck Re fills in an ACORD form with information such as the sewage conditions, flood zone information, and crime rate associated with the address. The system also pulls in data about the date of the building’s construction, as well as any details of remodeling done on the building and permits associated with it.
In addition, Planck Re claims it fills the ACORD form out with data such as floor material to assess the risk of someone slipping, recent health violations if the establishment sells food, and seating congestion if perhaps the building is a movie theater. The insurance agent would then be left with a completed ACORD form that they can then use to determine the business’ risk and make a decision as to whether or not to underwrite the business.
Planck Re’s software is not yet on the market. In fact, they started their series A round of funding in July 2018, raising $12 million from Israeli firms. As a result, they list no clients on their website, nor any case studies.
Their value proposition seems lofty, but we could infer it might work in a manner similar to Cape Analytics by leveraging machine vision. Alternatively, they could somehow plug into several databases containing the information they purport to be able to provide insurers, sifting through that scattered information to fill out an ACORD form for the insurer. Of course, this is speculation; the company does not provide any demonstration video of their service.
Elad Tsur is co-founder and CEO at Planck Re. He holds a Master of Science in computer science from Tel Aviv University. Previously, Tsur served as co-founder and CTO of BlueTail, which was acquired by Salesforce and was the basis of the company’s Sales Cloud platform.
Takeaway for Business Leaders in Insurance
It seems that AI-based insurance solutions are still in their infancy when it comes to claims processing and underwriting. Our research indicates that insurers looking for AI solutions may find more luck in claims processing technologies than in those for underwriting, but even then the companies offering claims processing software lack case studies and marquee clients.
Chatbot solutions appear to be the most mature claims processing technologies for insurers as of now, but only Lemonade claims to be able to automate claim payouts entirely. They are also their own insurance company, and they do not offer their software to other insurers.
Through our research, we found many press outlets covering Lemonade’s assertion that the company paid a customer’s claim in three seconds. In fact, the company has raised $180 million — far more than any other company listed in this report. Again, however, the company’s software is not a solution for other insurers.
Based on Elafris’ demonstration video, we could not determine if the company is actually leveraging machine learning in their chatbot software, and their executive teams’ lack of AI experience is a red flag. Tractable seems further along both in their funding and their use of AI.
The companies listed in this report claiming to provide solutions for the underwriting process are all startups without case studies or marquee customers. Planck Re is barely off the ground, and Daisy Intelligence focuses more on its retail solution.
Cape Analytics has raised the most money of the underwriting software companies we found. Based on its investors, the company seems to have some traction relative to the other startups listed in this report.
Insurers should expect chatbots for claims processing to become more available in the near term, perhaps due to the success of Lemonade. That said, complete claims automation does not seem to be readily available as a B2B solution.
AI for underwriting is a nascent space. More case studies are needed before we can say that the technology is viable for insurers at this time.
Emerj for Insurance Carriers
Insurance carriers come to Emerj for help assessing where AI fits in their business, including a map of which AI applications can best deliver ROI in claims and underwriting workflows. Emerj AI Opportunity Landscapes help insurance carriers pick first AI projects and select the right vendor for their particular business use-case, preventing them from losing thousands on pilot projects that are discontinued after only a few months. Contact us to learn more about how Emerj AI Opportunity Landscapes can help your insurance business win market share and compete with top players that are already leveraging AI to great effect.
Header image credit: Travelers Insurance
