Share this post:
Insurers know the importance of making informed decisions based on current, relevant information. Yet there are differing viewpoints on how much detailed information is necessary to make a decision. How much reliance can be placed on the “art” of underwriting or “experience” in lieu of gathering and reviewing a lot of detailed data and making decisions based on that evidence? How current does the information need to be – for example, when making a life underwriting decision, is it necessary to get blood test results (for medical underwriting) from samples drawn in the last 1-2 weeks, or can a three month old blood test result satisfy underwriting requirements?
Technology enables gathering vast amounts of data in real time (from sensors or IoT devices) for most types of property, life and health insurance risks. Technology also enables the analysis of both structured and unstructured data like handwriting, voice, video and social data. A significant challenge for businesses is the need to absorb and interpret increasingly large volumes of data before making necessary decisions, particularly as some decisions impact a growing ecosystem of business partners.
Quality of data is paramount
Although AI was viewed as experimental technology, more companies are now implementing AI-based solutions to rapidly analyze vast amounts of data, view trends, draw inferences based on predictive analytics, and to interact directly with customers or assist agents in responding to customer requests. However, the validity of the analysis and the AI-assisted responses to queries are dependent on the quantity of data examined, as well as the quality of the data. The old “garbage in, garbage out” adage is still true. When the results of analytics and AI can be relied upon, guesswork and risk in decision making can be mitigated. According to KPMG, 84% of CEOs are concerned about the quality of the data they’re basing decisions on. In a June 2017 interview with Jim Kramer on CNBC, IBM CEO Ginni Rometty stated “Twenty percent of the world’s data is searchable. Anybody can get to that 20 percent, but 80 percent of the world’s data, which is where I think the real gold is, whether it’s decades of underwriting, pricing, customer experience, risk in loans – that is all with our clients. You don’t want to share it. That is gold.” At the THINK 2018 conference in March 2018, she stated that by using AI systems like deep neural nets, they would be able to learn things about their businesses they wouldn’t have thought of on their own—and ultimately improve sales and ROI.
The art of intuition
Gathering information and curating data are two keys to effective decision management. However, the ability to operationalize data for decision making is the crucial capability insurers need to survive in a world of heightened competition and changing risks.
The “art” of insurance decision making comes from using learned associations and experience to account for missing data:
- For underwriters, using prior experience to account for unknown risk details has been central to decision making
- Actuaries have been conducting reviews of sampled data, developing correlating hypotheses and making decisions by analyzing several years of actual loss outcomes
- When settling claims, adjusters predict likely outcomes based upon their experience, which is employed to reduce claim costs and avoid unexpectedly large settlement amounts.
More nuanced decision making
Cognitive systems can help agents anticipate and improve customer experiences. For insurance decision makers, these systems can help eliminate routine, rote work, which gives decision makers more time to focus on complex tasks that require creativity and empathy. Better and more current data is also giving companies the opportunity to develop new business models and create new products.
Human aids, not humanoids
In a 2016 Institute for Business Value (IBV) survey of more than 1,500 C-suite insurance executives, 56% said reducing underwriting risk was a high business priority, second only to improving operational efficiency.
I recommend considering two of the decision-making processes in operational insurance. First, actuarial assessments involve loss development trending and pricing projections. Actuarial development of aggregated information into data segments can be used for predictability. Product pricing can then be developed that responds to that predictability. Second, underwriting program decisions are initially informed by the actuarial guidance on risk classification. The pricing is then refined based on any extra data or information that provides an underwriter with a clearer sense for the risk. Better decisions can be made by following a consistently accessible way to deliver extra data to the decision maker.
AI and cognitive computing relate to businesses and customers in four ways:
- Understanding: Cognitive systems understand images, natural language and other unstructured data.
- Reasoning: Cognitive systems can reason, grasp underlying concepts, form hypotheses, and infer and extract ideas.
- Learning: Cognitive systems learn from experience. When assessing a risk or a claim, the ability of some cognitive systems to read the equivalent of 800 million pages per second gives them unprecedented capabilities to review similar cases and make better recommendations.
- Interacting: Cognitive systems interact with humans using human language, not code.
Enterprise data governance
To help avoid missteps, the best implementations of cognitive decision making also improve how data is governed, data architecture is used and how related solutions are implemented Improving how a company manages decision making isn’t as simple as throwing a switch. It requires broad agreement on the “sources of truth” from which reliable data is gathered.
To begin the journey to better decision making, there are clear steps insurers can take. To learn more, please read the IBM Expert Insights Paper “Cognitive decision making in insurance: From art to science.”
Learn more about IBM Insurance solutions here.