For a long time we've understood that computers can do certain things better than humans. Math instantly comes to mind, as do repetitive tasks that don’t require creativity or advanced problem solving.
Knowing just what humans are good at and actually want to do versus what computers and artificial intelligence (AI) should perform is the next frontier for many industries, including insurance. AI will make our jobs and lives easier, if not richer. We’ll focus on what actually matters instead of what has to get done.
We are already seeing this take place with AI and computer vision transforming property and casualty insurance claims. AI already has three core value adds that will transform the insurance industry. But first, let's examine what I mean by AI for insurance.
What is computer vision?
For a long time, only human vision could truly perform expert tasks, such as evaluating an accident.
These days, however, AI is making sense of imagery using computer vision — an algorithmic approach to gaining a high-level understanding from digital images or videos.
From an engineering perspective, computer vision seeks to automate tasks that the human visual system can do. The field has seen major advances that allow machines to see in complex ways and we are at a point where many visual tasks can be performed more efficiently by machines than by humans. Driving is an obvious one: Americans spend an average of 300 hours driving per year. AI has been causing a great stir in the automotive industry with autonomous vehicles. Companies like General Motors and Ford, that 10 years ago we’d think would have nothing to do with software, have both invested more than $1 billion in AI.
AI meets insurance
What is happening with AI and the automotive industry will soon happen to the property and casualty insurance industry because visual tasks prompt most claim management decisions.
Insurance is about protecting things of value: human life (life insurance), human health (health insurance), and human property (property and casualty insurance). Property and casualty insurance is a $540 billion industry in the U.S., and 70 percent of the spend is on claims. A property and casualty claim occurs when there is damage to a piece of property, be it a car, a house or infrastructure.
Diagnosing and treating damage is determined visually. Take, for example, a car that has suffered a hailstorm. To treat this car, we need to know:
— Which parts are damaged?
— What do the part require to be fixed (repair, replacement, paint, etc.)?
— How much labor time is required to perform each repair?
All of these decisions are determined visually. If the total repair cost is greater than the vehicle value, the car is a totaled. Vehicle salvage value is also determined visually.
What if AI were to make these decisions (remember, computers are better at math and can now see well too)? Sounds exciting, but what benefits would there be to property and casualty insurers?
Automating the analysis of visual data can bring value in three different ways: accuracy, speed and scale...
Property and casualty insurance is a $540 billion industry in the U.S., and 70 percent of the spend is on claims. (Photo: iStock)
Value add No. 1: AI accuracy
A major benefit of AI and computer vision is analyzing data more accurately than humans, and making fewer mistakes. Computers can be made and trained to be more consistent and exhibit less bias than humans. They also don’t get distracted, tired or hungry, which are factors we’ve all dealt with when taking on tasks.
Here’s an example: Bodyshops and insurers do not always agree on the best way to repair a car. Biases can arise from the economic incentives of both parties: a bodyshop will typically benefit from a high repair cost, while an insurer will typically benefit from a low one. This can often result in back-and-forth between insurers and shops as they argue over the right procedure, which in the end adds inefficiency costs that delay claim settlement.
AI that has been trained to apply unbiased repair standards, mutually agreed on by both insurers and bodyshops, can instantly apply unbiased standards, remove inefficiency costs and speed up claim settlement. Images taken by bodyshops or insurers can be uploaded to the AI, from which an unbiased repair process can be created — thereby driving socially optimal outcomes. This is already being used by insurers around the world, such as Ageas.
A major benefit of AI and computer vision is analyzing data more accurately than humans, and making fewer mistakes. (Photo: iStock)
Value add No. 2: AI scale
There are only so many hours in the day and another value add of AI is its ability to analyze massive quantities of data faster than ever before. This matters when there is too much data for humans to exhaustively analyze and therefore only a fraction of what should be analyzed gets the expert attention it deserves. However, now AI is coming in to enable full scale analysis for the first time.
This leads to many more discoveries that could have enumerable benefits for insurance companies.
Another way to think about it is that there are 25 million auto insurance estimates per year in the US alone. These estimates contain repair processes and images of the car damage to back up the claim. It is advantageous to check these estimates when they come in to verify that repair recommendations are accurate. This helps insurers ensure that quality and efficiency standards are met. But these checks require human expert judgement. This can be slow, expensive and difficult to operationalize.
Today, approximately 45 percent (11 million) of vehicle estimates in the U.S. are not visually checked because it is not economical to do so. This gap equates to $27 billion of repairs. AI can make a dent in this huge opportunity by flagging useful cases to humans. Instead of humans inspecting all estimates and getting a 15-20% hit rate, they inspect those suggested by the AI and get a 70-90% hit rate (based on how well AI is calibrated to their standards).
AI also organizes the information efficiently (showing an appraiser specifically which images to look at), which helps humans go up to 5 times faster. Overall productivity (aka hit rate and review speed) increases by up to 20 times. Now it makes economic sense to review more claims, increasing the need for human labor because there could be up to 3.3 million new insights. The good news is that in this use case, AI would not replace jobs rather potentially create up to 3,000 new jobs in the US alone.
By our calculations at Tractable, AI and the necessary humans could save Americans up to $1 billion in processed claims.
Tractable has determined that AI could save Americans up to $1 billion in processed claims. (Photo: iStock)
Value add No. 3: AI speed
Last but not least, AI brings value by analyzing data earlier on in the claim process, typically seconds after the data is captured versus days when inscribed via a manual process. This gets interesting when the time difference allows you to act before an adverse event occurs.
For example, you could analyze car photos as soon as a customer reports a claim and uploads images, instead of several days later. This allows insurers to determine right away if a car is a total loss and therefore avoids unnecessary towing and parking of the vehicle. You also avoid customer dismay (the terrible experience when a customer drives the car to the shop for an estimate, drives it back home only then to have the insurer say the vehicle is totaled). Along these lines, insurers could determine the cost of repair and propose a cash settlement on the spot.
This is the end goal of AI for property and casualty claims: enable same-day, fully-automated claim settlement.
Despite being a new technological approach to the property and casualty claims process, AI and computer vision are already proving their value with these value-adds. AI offers new, streamlined workflows that make the claims process better for insurers and customers alike.