top of page
  • Writer's picturejason

Rules Engines, Predictive Models and AI. Where to start?

Working in the life insurance industry has got a whole lot more interesting over the last couple of years. Of course, some might argue this is not new. As an underwriter, it was always fascinating to review medical histories, assessing applicants with unusual sports and pastimes and dangerous occupations. However over recent years, the availability of detailed, high quality data, increasingly accessible in real time, has changed the game. Initially this data was just a way to speed up the process, but it’s ultimate impact is to bring a different way to assess risk.

We are now surrounded by a range of tools and approaches that offer the chance to fully modernise our business. Most insurance carriers have, at least, experimented with some of these solutions, but many are still waiting on the sidelines. In an age of digital transformation, its hard to believe that the majority of life insurance applications in North America continue to be submitted in paper. Whilst the end of paper has been prophesied for decades, old habits die hard.

However, the pace of change is increasing and its time for companies to embrace the new, or prepare for an unrecoverable slide into irrelevance. Assuming you buy into this premise, the key question becomes, where do you start? Over 50% of carriers have now deployed (or are deploying) underwriting rule engines, but some are questioning whether these will be superseded by predictive models and AI.

Rule engines offer a more efficient and consistent way to underwrite, but the principle remains that each case is assessed individually. The advent of big data offers an aggregated alternative, based on statistical models that predict mortality using proxy data: bank, credit, buying behaviour etc. Some see this as an either/or option. I visited a company a couple of years ago that openly admitted they were having a battle at a senior level over which approach would be prevail in the long term. My answer was that there is probably not a single answer to the question, and that both approaches have a role. In general, this is how things have played out, with predictive models mostly used to triage underwriting requirements and rules engines / traditional underwriting providing the final decision. Of course, this continues to be an area of much activity, not just from carriers but providers and regulators, so it will be interesting to see if this balance remains.

AI is worthy of a separate discussion, but will undoubtedly start to have an increasing impact moving forward. Some companies have ventured into this space early, but with generally disappointing results. However, the advent of electronic health records (EHR) is going to accelerate the importance of cognitive computing. Considering an APS statement is often hundreds of pages long, the volume of data generated by EHR will be beyond the scope of the manual process, and will even overstretch the capabilities of a rules-based solution. A different approach will be needed, that is capable of compressing and pre-assessing data for final decision by the expert underwriter. We are already seeing AI tools perform this type of role in the medical profession e.g. assessing cancer patients data and recommending treatments for a doctor to prescribe. This will be a fascinating area of development but will require significant investment across the industry.

Overall there is a lot of opportunity in these areas and this should ultimately lead to a better result for ours customers (hopefully helping reverse the decline in life insurance sales). It looks like interesting and exciting times ahead.

27 views0 comments

Recent Posts

See All

Project pitfalls and how to avoid them

I had cause to reflect on failure this week, following a homemade pizza disaster, that my teenage daughter described as an “epic fail”. There were definitely a series of unfortunate events that led to


bottom of page