Simply put, AI provides powerful support to decision-making without displacing the decision-makers. While some decisions may be effectively automated based on the presence of highly predictive factors, the greatest value is derived from allowing decision-makers to devote most of their time and talents to complex choices where their expertise will make the most difference.
The Agent-Carrier Partnership
For machine learning to be truly successful, its implementation must be grounded in a business context with a well-thought-out series of problems to solve.
In life insurance, agents are an indispensable source of observations on market needs and characteristics. From agents’ input, carrier analysts and data scientists determine what data is needed to build predictive models for identifying different types of risks and the underwriting and pricing needed to make them sustainable classes of business. In a recurring loop, agents then assess how precise a model or models prove to be in predicting loss experience, and carrier staff refines the model to make it even more precise.
Today, there are numerous applications of AI and machine learning in life insurance, including:
- Marketing optimization, which uses models to identify prime market segments and the best methods for reaching the right prospects. This helps agents allocate their marketing expenditures and outreach efforts most effectively.
- Lead and prospect evaluation, done through models that score individual prospects to predict how likely they are to buy a product, the breadth of products they might purchase, and their likely tenure as a customer. This helps agents prioritize which leads to pursue first.
- Pricing and risk assessment, done through models that evaluate the relative risk that each prospective customer represents, drawing the agent’s and underwriter’s attention to key considerations in evaluating and managing the risk. This is crucial for identifying a competitive price that matches the risk level of the customer.
Data and Privacy
In theory, anyone, including life insurance agents, can build their own machine learning models. But it will come as no surprise to hear that AI works best with lots of data capturing lots of variables — literally millions of records — more than most organizations can obtain or afford.
This demand for data creates the great expectation that organizations and their staff members will implement sound data governance and strong privacy protections, especially in a field like life insurance where practitioners are guardians of information on personal health conditions. You might possess health information that could be useful in identifying additional products a customer could use, but due to restrictions on the sharing of health-related information, you are not always allowed to use the information for that end.
A relaxation of privacy standards could help life insurers gather information more quickly and proactively identify customer needs and demand. For now, however, let’s make the most of what we have to understand the life stage of each of our customers and prospects and to identify and measure their needs as they age.
— Connect with ThinkAdvisor Life/Health on Facebook, LinkedIn and Twitter.
Neal Silbert is general manager, insurance, at DataRobot, a Boston-based company that offers an automated machine learning platform. He previously was vice president of predictive analytics at Zurich North America.