Artificial Intelligence: The Next Battleground Kevin Mason, EVP Solutions | Apr 03, 2019

Insurance Artificial Intelligence

There is a lot of buzz about artificial intelligence these days. From science fiction writers to movie makers to celebrities, everyone has a point of view on AI and it’s typically pessimistic. Oxford philosopher Nick Bostrom is one of those. He says, “we humans are like small children playing with a bomb. We have little idea when the detonation will occur, though if we hold the device to our ear we can hear a faint ticking sound.”)1 Not exactly a cheery prognosis.

Whatever your view of AI, you can’t ignore that fact that it’s fast becoming the next big battleground for business and government. So, if I am an insurer, how do I use it to compete in the Darwinian Economy? In this post, we’ll explain what AI is, how it applies to insurers large and small, and steps you can take to get started.

AI and Machine Learning Explained

AI applies advanced algorithms to digital data, extending the human ability to sense, learn, and understand.2 Microsoft breaks AI down into three stacks: Cognitive Services, Robot Process Automation (RPA), and Machine Learning (ML). In this post, we will unpack machine learning.

Machine learning is the application of algorithms and statistical models by computer systems to perform a specific task without explicit instructions, relying on patterns and inference instead.3 The goal of machine learning programs is to identify cause and effect relationships that explain past behavior and predict future behavior. If I know, for example, that as a business owner you exhibit signs of needing D&O insurance, I can create a marketing campaign that will direct you to my product. 

The fuel that feeds machine learning is data. Data is used to train machine learning algorithms. In simple terms, machine learning takes sample data called “training data” and builds models from it. The models are then used to predict future behavior and make decisions without the need to have been programmed to do so. The better the data, the better the decision making.

No better example exists than in the game of Go, one of the most complex strategy games, and once considered the last line of defense against machine learning algorithms. In 2016, a computer program beat one of the best Go players in the world. Overnight, humans were no longer the best Go players on the planet.4

Insurers are using machine learning in lapse management, recommendation engines, assessor assistants, property analysis, fraud detection, personalized offers, experience studies, and scaled training.5 As the number of use cases expand, insurers who do not have this capability will lose out to competitors.

AI Levels the Playing Field

On the AI battlefield, small and large companies will compete on equal footing. In an interesting twist, size doesn’t matter (except for data, which we discuss later). For starters, most AI environments are cloud based, so even the smallest company can tap into a scalable hardware environment at a fraction of the cost large companies spend on data centers.

Data scientists, likewise, have shown no preference between large and small companies. What matters to them is the nature of the work and the freedom they are allowed. So, small companies have the same access to talent that large ones do. In other words, the AI battlefield negates size as an advantage.

With Data, Size Does Matter

The ammunition used on the AI battlefield is data. Here, size matters. The more data obtained, the more likely results will be statistically significant. (With small data amounts, results can be skewed significantly by one or two samples). For insurers on the commercial side, the small business market will most likely yield the greatest treasure trove of data while steadily decreasing as you move up to mid-size and large commercial accounts.

In all cases, insurers will use historical data to predict future behavior. For those with small data sets, the challenge will be to identify cause and effect correctly. Correlation does not equal causation. The trick with small sample sizes is to use the right methods to identify patterns that tell you something valid. Pulling in third party data sets that provide different perspectives is one way to do this.

Getting Started

Machine learning is capable of building algorithms that will identify the correct product to sell to the buyer at just the right time. Insurers can do four things to make sure they are not left behind.

  1. Educate yourself on machine learning technologies. Understand the landscape of the arena you will be battling in.
  2. Begin small pilot projects within the company using whatever data you can find. The goal of this step is to gain experience, not create insights. You want to learn as much as possible so embrace failure and iterate.
  3. With the learnings of steps 1 and 2, identify projects that can be kicked off within your markets and launch them. Some will succeed, some will not, but in all cases, you glean information from the projects and get better and better at understanding machine learning.
  4. Use your learnings to identify patterns and opportunities in your product portfolio that will benefit from the use of machine learning.

Machine learning is the base skill for AI. In our next post, we will talk about two additional layers that make heavy use of machine learning – cognitive services and robot process automation.

Sources

1 Adams, Tim. Artificial intelligence: ‘We’re like children playing with a bomb’. The Guardian. June 12, 2016.
2 Artificial Intelligence Research. Retrieved from https://www.microsoft.com/en-us/research/research-area/artificial-intelligence/
3 Machine learning. Retrieved from https://en.wikipedia.org/wiki/Machine_learning. April 1, 2019.

4 Borowiec, Steven. AlphaGo seals 4-1 victory over Go grandmaster Lee Sedol. The Guardian. March 15, 2016.
5 Seely, Scott. Eight use cases for machine learning in insurance. Azure.microsoft.com. October 2, 2018.

Kevin Mason has worked in many aspects of software development since 1981, including roles as product strategist, software development methodologist, project manager, and technology architect for companies such as Cincinnati Bell Information Systems, SHL Systemhouse (now part of EDS), AGENCY.COM, and GENECA. He joined Instec in 2008 and is responsible for development associated with all products. He holds a BA in Political Science, from the University of Iowa and an AS in Computer Science.