AI in Auto Insurance – Current Applications

Marcus Roth

Marcus Roth is Operations Manager at Emerj. He manages content and marketing processing, and helps with research into Emerj's primary business sectors.

AI in Auto Insurance - Current Applications

KPMG estimated the size of the automotive insurance is expected to shrink by 70% due to the rise in demand for autonomous cars and the shift in liability then being placed on the car manufacturer. With the rise of AI in most sectors, it follows that AI would find its way into the automotive insurance world. AI may allow car insurance companies to keep up with an evolving consumer base that is looking for faster service, faster payouts, and policy prices tailored to them.

Recently, we published a report looking at the AI initiatives of the 4 largest insurance companies in the United States. This report seeks to cover the AI initiatives of the auto insurance sector, specifically. As of now, numerous companies claim to assist auto insurances in aspects of claims processing to handling customer support.

We researched the space to better understand where AI comes into play in the auto insurance industry and to answer the following questions:

  • What types of AI applications are currently in use in auto insurance?
  • What tangible results has AI driven in auto insurance?
  • Are there any common trends among these innovation efforts? How could these trends affect the future of auto insurance?

This report covers vendors offering software across four applications within the auto insurance industry:

This article intends to provide business leaders in the auto 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 auto 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 auto insurance spend researching AI companies with whom they may (or may not) be interested in working.

Claims Assessment

Ding Sun Bao

Ant Financial, a company part of the Alibaba group, is a Chinese fintech company with over 3,000 employees. The company created a software called Ding Sun Bao, which it claims analyzes vehicle damage and handles claims using machine vision.

Ant Financial claims a user can take a photo of their vehicle damage using their smartphone camera or upload photos of the damage into the software. Then, Ding Sun Bao compares the uploaded image of the damage to a database of images labeled as various severities of damage. These images might also be labeled with likely repair costs. Then the application produces a report for the user on the damaged parts, a repair plan, and the accident’s impact on the user’s premiums in the years following the accident.

We were unable to find a video demonstrating how Ding Sun Bao operates.

Ant Financial claims it put its AI up against six insurance claims specialists and compared them on the speed with which they handled the claim. According to the company, the AI “assessed” the damages and handled the claims in six seconds. The human claims adjusters apparently took six minutes and 48 seconds to reach their conclusions.

Ant Financial claims to have helped China Taiping, China Continent Insurance, Sunshine Insurance Group, and AXA Tianping process claims faster than their human adjusters. Yin Ming, President of Ant Financial, claims “Dingsunbao has already helped the insurance industry save over 1 billion Chinese Yuan on claims handling while saving claims adjusters around 750,000 hours of effort.” That said, Ant Financial does not provide any statistics reporting the results the software might have generated for clients in the form of a case study.

Li Cheng is CTO at Ant Financial. He holds a Master’s degree in Computer Science from Shanghai University.


We previously covered Tractable’s software for aiding claims agents in our report AI for Claims Processing and Underwriting in Insurance.

Tractable is a UK-based company that offers a software which it claims can help insurance agencies automate the claims process 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 photographed damage. Then, Tractable’s AI compares the uploaded image to a database of various images labeled with varying 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 so, it displays a warning to the agent to decrease the payout. As a result, the software can reduce the amount of excess money distributed to a client when agents pay claims.

In the video below, Tractable’s Chief Operations Officer and co-founder, 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.

Razvan Ranca holds a Master’s Degree in Computer Science from the University of Cambridge and is CTO at Tractable. 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.

Chatbots for Customer Service


Geico offers a virtual assistant and chatbot named Kate, which it claims can help customers answer questions they have about their auto insurance using what seems to be natural language processing.

Geico claims users of their customer service app can message or speak to the Kate software to inquire about policy coverages, view billing information, and direct them to the appropriate section of their insurance application. It is unclear how the AI was trained to receive that input from the customer. In other words, we were not able to concretely find information on how the software takes user input and turns it into a text response to the user.

Given that it is able to appropriately handle a variety of insurance policy questions by the user, it may be inferred Kate’s decision-making ability and her ability to process what is spoken or written to her is natural language processing. Potentially detecting the use of certain words or phrases, such as “my policy” and “mechanics I can go to” and generating an appropriate response in the form of a pre-written message or pre-formulated verbal response.

We can infer the machine learning model behind the software was trained on hundreds of thousands of customer support questions and thousands of snippets of recorded speech taken from Geico’s normal methods of human customer service. This text and audio data would have been labeled as a certain type of question, such as “refund request” or “policy question.” The labeled text and audio data would then be run through the software’s machine learning algorithm. This would have trained the algorithm to discern the chains of text that, to the human brain, might be interpreted as a question about one’s auto insurance policy as represented in a snippet of speech or text.

A demonstration video for Kate could not be found.

Business leaders should note that by using Kate, Geico might be able to free up human customer support agent time, allowing less resources to be devoted to answering relatively simple policy questions and allowing for more time to be spent by the representative elsewhere.


Progressive offers a chatbot called Flo, which the company claims can help customers using natural language processing and cloud-based API insurance data to alter payment schedules, file insurance claims, and request auto insurance quotes.

Progressive claims Facebook users can interact with Flo utilizing Facebook Messenger. Users can message the Flow software insurance questions regarding filing claims, billing, and premium rates. The user’s message is run through the machine learning algorithm behind the chatbot. The chatbot might know whether or not the message is a claims or billing question because it would have been trained on thousands of text-based customer support messages labeled as claims or billing questions.

The system then provides the user with a text reply that might provide them information about the status of their claim or their next invoice. These responses would have at first been pre-written templates that the chatbot was programmed to send upon receiving a message of a certain type.

Progressive uses Flo internally, and so there are no available case studies for the software. That said, Microsoft Azure claims to have helped Progressive build their chatbot, which emulates their popular TV mascot’s key mannerisms in addition to “understanding” customer questions and responding quickly. Flo attempts to engage customers with commentary that ties into the marketing and commercial advertisements it runs.

Progressive built the Flo Chatbot using Microsoft Azure Bot Service and LUIS. The company found that the software facilitates updating the bot and its responses without needing to write complex code in-house. Microsoft reported that Progressive updated Flo’s models over 75 times in the first four months of activity to help the chatbot continue to improve itself and customer interactions.

There does not seem to be any C-level executives on the team with an AI background. Progressive is an industry leader in auto insurance, but we caution readers to be wary of companies that claim to do AI without any C-level AI experts on their team.

Policy Pricing

Ant Financial

Ant Financial also offers an AI software called Auto Insurance Points. Ant Financial claims Auto Insurance Points can help auto insurance companies produce more accurate customer risk profiles and appropriate pricing using machine learning.

Business leaders may notice that Ant Financial seems to be a big player in AI as it pertains to the insurance industry. The MIT Technology Review added Ant Financial to its list of the 50 Smartest Companies of 2017.

Ant Financial claims insurance companies can enter policyholder data into their software. Auto Insurance Points consider regular factors as the year and car make and model, as well as less traditional data points such as the policy holder’s profession, credit history, spending habits, and driving habits. Then the software provides a score between 300 to 700, where 700 means the driver is low risk. A more accurate depiction of the customer and their lifestyle could allow insurance companies to offer more tailored rates to customers, allowing low-risk drivers to receive discounts and allowing the insurance company staying competitive.

Much like Ding Sun Bao, we were unable to find a video demonstrating this software.

Ant Financial does not list any major companies as clients for their Auto Insurance Points software. The business application for this software being most applicable to small and medium-sized insurance firms as they may be able to compensate for the comparative lack of risk assessment staff with the software able to process many of those risk factors for them and lend aid in offering the most strategic price possible.

What Business Leaders in Auto Insurance Should Know

Many older companies are starting to hop on the AI bandwagon and use AI as a buzzword to generate interest in their brand. In many cases, large enterprises

Insurance expertise does not inherently make a company an expert on the effective execution of artificial intelligence.

It would seem the automotive insurance sector as a whole is just beginning to adopt artificial intelligence. In fact, some of the larger automotive insurance companies are outwardly resisting artificial intelligence in the space. State Farm recently released a commercial seemingly targeted at a disruptive insurance company called Lemonade, Progressive’s Flo, or Geico’s Kate and the fact that they claim to use AI as a core selling point. In the commercial, the actor playing the State Farm representative mocks the robotic claims agent representing its competitors and highlights that State Farm employs 19,000 human agents.

This contention may highlight the uncertainty many have when utilizing AI. To those companies not prepared to adopt AI into their business model, due to its large development cost or otherwise, AI could appear as a large waste of company resources. In terms of budget, State Farm may not be ready to commit to an AI initiative, and so a solution to avoid that cost while attempting to still appear aware of market trends is to acknowledge AI, as they have, and extol the virtues and classic comfortability of human customers dealing with human customer service representatives.

Many auto insurance companies claim that a large pain point for their customers occurs when filing claims. Insurance companies report that negative claims experiences will cause customers to change insurance providers. Additionally, the claims process causes the largest number of negative experiences, so it would seem to behoove leadership to find solutions that would make the claims process as smooth and enjoyable as possible. AI may prove to be a more scalable way to reduce friction through instant customer support via chatbots and faster claims assessments through software such as Tractable or Ant’s Ding Sun Bao.

The CEO of the world’s largest telematics company, Octo Telemetrics, Fabio Sbianchi, recently made a statement regarding AI seeming to suggest the automotive industry will be looking to AI to solve its problems in some way:

“Nowadays, we are able to use the best software and strategies, that are based on machine learning and AI, for the analysis of #telematics data. We have moved from the collection and analysis of static data to that of dynamic data, which represent a significant change in the standard insurance model.”

State Farm, Progressive, and Allstate all utilize IoT telematics programs that allow users to install telematics devices in their cars for discounts. Theoretically, the more telemetric data is strategically utilized per client the more specific the pricing and policy decisions can be for insurance companies. It should be noted that it doesn’t seem as though these telematics programs are using AI.

That said, the data that they collect for these companies can and likely will be used to train machine learning algorithms. In doing so, companies might be able to better predict the risk of its customers and offer tailored pricing to them, similar to what Ant Financial is doing. If they can do this, they can theoretically beat out their competitors to the most valuable customers.

Header Image Credit: Hayes Insurance Brokers