By Mark Smith, President of the Global Insurance Practice
That is all the time it took for AlphaZero—the chess-playing AI created by Google’s DeepMind—to learn the rules of a game human beings had dominated for fifteen hundred years…and then vanquish the world champion chess program in a one-hundred-game matchup.
“It doesn’t play like a human, and it doesn’t play like a program,” DeepMind’s Demis Hassabis, a child chess prodigy himself, said at the 2017 Neural Information Processing Systems (NIPS) conference in Long Beach. “It plays in a third, almost alien, way.”
“We have always assumed that chess required too much empirical knowledge for a machine to play so well from scratch,” former world chess champion Garry Kasparov told Chess.com, “with no human knowledge added at all.”
How did it happen?
MIT Technology Review has the details:
“What’s also remarkable, though, Hassabis explained, is that it sometimes makes seemingly crazy sacrifices, like offering up a bishop and queen to exploit a positional advantage that led to victory. Such sacrifices of high-value pieces are normally rare. In another case, the program moved its queen to the corner of the board, a very bizarre trick with a surprising positional value. ‘It’s like chess from another dimension,’ Hassabis said.
“Hassabis speculates that because Alpha Zero teaches itself, it benefits from not following the usual approach of assigning value to pieces and trying to minimize losses. ‘Maybe our conception of chess has been too limited,’ he said. ‘It could be an important moment for chess. We can graft it into our own play.’”
So, if our “play” is insurance, can we “graft” the lessons of AlphaZero onto it?
Absolutely.
The AI Revolution in Insurance
Artificial Intelligence—which applies computer science to extensive datasets in order to accelerate problem-solving and power search engines, product recommendations, and conversational AI systems—has proven to be a key enabler of digital transformation in insurance.
In the case of, say, a digital virtual agent fielding some portion of the claims process—billing assistance, claims status inquiries, auxiliary call center staffing for roadside assistance requests—this version of AI is in practice probably not all that different than the version Hollywood has been presenting for decades now. In most areas of insurance, however, AI is dovetailed with robotic process automation (RPA) to establish intelligent process automation or, more commonly, intelligent automation.
Typical applications include machine learning and deep learning to make predictions based on input data—e.g., claims and underwriting data. In addition to increasing efficiency for repeatable processes, this workflow automation plays a significant role in detecting potential fraud and enhancing the customer experience.
Today’s insurance industry faces an ongoing challenge in delivering profitable and predictable results. Most of the industry’s global equity had a return on equity (ROE)
below the cost of equity over the past five years.
Given the industry’s ongoing financial challenges, forward-looking leadership are turning to AI to achieve savings and transformation.
Naturally, some of this acceleration is hurtling ahead in the tailwind of the pandemic: A PwC survey found more than half of respondents had hit the gas on AI adoption plans due to COVID. Further, a full eighty-six percent believe AI will be a “mainstream technology at their company” in the very near term.
That is already an unusually wide consensus—yet it is poised to grow even larger.
A recent McKinsey study estimates the total potential annual value of AI adoption to the insurance industry alone at $1.1 trillion dollars. That is a staggering—perhaps unprecedented—ROI. And the benefits are hardly siloed.
Digging Deeper in DeepMind’s Success
The difference between AlphaZero and its competitors, according to DeepMind, is that its machine-learning approach was provided with no human input apart from the basic rules of
chess. The rest works out by playing itself over and over with self-reinforced knowledge. The result, according to DeepMind, is that AlphaZero took an “arguably more human-like approach” to the search for moves.
Consider an auto accident. The process begins with “first notice of loss,” FNOL, the awkwardly named start to claim processing and adjudication. FNOL is a well-defined framework with detailed process flows, regulatory requirements, and multiple touchpoints for the customer. For example, customers can call their agent, use the toll-free number for the call center, use the app on their phone or tablet, or log into the insurance company’s site.
Multiple touch points represent one set of flows. There are also numerous scenarios depending on whether the car was damaged, the person calling sustained injuries (first-party), or the other driver was hurt (third-party). There are many areas where Insurers can automate the process to increase efficiency, lower expenses, and even catch signals of potential fraud.
Much has been done to apply AI and machine learning to streamline claims processing. Yet FNOL is much more than an exercise in efficient operations. Dealing with a claim is among the most human of experiences because it begins with emotional loss. A person has been in an accident. They’re unsettled, very likely still at the scene of the accident. They are asked a series of questions by an insurance claims representative.
Consider the claims process for an auto accident. A person has been in an accident. They’re unsettled, very likely still at the scene of the accident. They are asked a series of questions by an insurance claims representative.
Two days pass. The customer receives a call from a different insurance company claims representative. They are asked additional questions, likely repeating a few questions asked at the accident site. They’re asked to submit another form or two. Presumably, they have questions, too, including which auto body shop to use or how to submit medical bills.
Enter the power of conversational AI. Paradoxically less technical and more intuitive, conversational AI lends itself to a human approach. Think fast-evolving Alexa, not the static IVR (Interactive Voice Response) technology that we all dread:
Press one for sales, two for service, three for…
Conversational AI is moving rapidly to change the claimant’s experience. Looking forward, it’s going to accelerate even faster. Google’s DeepMind didn’t stop at chess. Its recent iteration, MuZero, uses “deep reinforcement learning” to let machines teach themselves new skills via a process of trial and error, receiving “rewards” for success rather than being told what to do. Still in development, potential uses included next-generation virtual assistants that could support customer service through conversational AI and more.
Streamlined processing, enhanced fraud detection, and superior claims service are just a few examples of AI applications driving Intelligent Automation—and perhaps ensuring its very survival.
OZ partners with insurers to leverage AI & Intelligent workflow solutions for Claims, Underwriting, submissions, call center, and renewal workflows.
Contact Mark Smith, President of the Global Insurance Practice and Insurance Practice Leader, to discuss how Intelligent Automation can streamline processing and enhance profitability for your business.