Data is at the heart of artificial intelligence and machine learning projects in the insurance industry. The more data insurers have, the better their results. (Shutterstock)
The COVID-19 pandemic has impacted every industry around the globe, but perhaps few as much as the property & casualty insurance industry. From an incredible increase in claims payouts to fears that the economic slowdown will result in customer loss, the industry is struggling to overcome multiple challenges while continuing to meet the current and future needs of its customers.
The insurance industry’s biggest and most valuable resource is the data that insurance companies have collected over the years on everything from weather risk to consumer behavior. Insurers had only just begun to incorporate artificial intelligence and machine learning (AI/ML) into processing this data before the pandemic erupted.
Now, new pressures are making it even harder for larger insurers to keep pace, and even more difficult for mid-sized and smaller insurers to adopt AI/ML technologies because of long development timelines and traditionally high investment requirements. As such, many of these companies are looking to innovative new technologies to help optimize their AI/ML investments and/or launch AI/ML programs with limited resources.
COVID-19 and insurance: A closer look
A recent dotData survey conducted with a variety of property & casualty insurers revealed several surprising facts and trends about the impact that COVID-19 has had on AI/ML deployment in the insurance industry, and how insurers are looking to mitigate that impact. These include:
Further follow-up questions helped illuminate specific pain points and goals along with ways insurers can better use AI/ML to drive business impact. These strategies include:
No. 1: Understanding your data
Data is, of course, at the heart of AI/ML projects. The more you have, the better the results will be. However, the pandemic has completely changed what types of data are needed for a successful project. There are new data sets and requirements that must be incorporated into efficient models, which makes their creation more complicated.
Knowing the type of data you need and already have makes AI/ML projects successful. And it’s just as important to know where your data is stored and how each set impacts your business. Mapping out how to gather data and consolidate it into other models takes time up-front. But doing so results in better, more accurate models processed in a faster time.
No. 2: The need for speed and agility
A recurring theme in our study showed that models need to have greater detail to help executives make decisions. This was true before but the pandemic changing the types of data needed has emphasized this point. Further, many insurers have also stressed that they have an increased need to integrate data from external sources like government sources into their models and systems.
Because of this, insurers have said that the issue of speed and flexibility in AI/ML systems was badly needed. The best way to handle this is to incorporate this factor into your processes from the beginning. Take a look at what processes are the most time consuming. Find out where bottlenecks exist. And research the best ways to adopt new tools into your platform to help you find answers and solutions quickly and efficiently.
No. 3: Involve customers early
Any data you use is there to ultimately help you find the best ways to serve your customers. Insurers listen to their needs and concerns when issuing and underwriting policies. Insurers who underwrite policies protecting hospitals and other medical facilities against liability claims need statistics and numbers as specific as possible. Some that come to mind: How many patients and employees flow in and out of a building? Historically, what are the types of incidents that result in liability claims and what can be done to lessen those risks?
These types of specific concerns should be incorporated into any predictive analytics project. This means having the tools at hand right from the start to know what your customers want to know. Also, make sure you can simply explain your findings to your customers, which shows that you know how to “connect the dots” and engender the trust needed for a positive, lengthy business relationship.
No. 4: Develop, test, rinse and repeat
Data science modelling is predicated around the ability to make assumptions, test those assumptions and then make appropriate changes. But it takes a lot of time, effort and ultimately money to build, evaluate and then adjust models. Insurers should thus have the tools in place to accelerate and automate as much of this process as possible. AI/ML doesn’t just make harvesting and analyzing data faster and more accurate. AI/ML also lowers costs down the line while also freeing up limited resources for other uses.
No. 5: Not a legacy anyone wants
The pandemic has shown where many flaws and cracks of our world lay bare. One major issue comes in the form of “legacy” computing systems that were programmed in now outdated systems. This issue has plagued a number of local, state and national government websites since COVID-19 entered our lives.
It has also been incredibly problematic for many in the property & casualty insurance industry. Multiple companies reported they still have systems originally written in COBOL, which was designed decades ago and no longer taught in most computer science programs. This means that many insurers rely on computing systems built a long time ago at the absolute moment they must step into the future. Companies that have invested in new core computing systems have been able to incorporate AI/ML into their data analysis quickly and affordably. Firms that are still trying to patch together decades-old systems with modern data processing technologies find themselves struggling to adapt in a world that demands it.
AI/ML was already reshaping how insurance companies incorporate data into their decision making and strategic processes. Due to the unprecedented era we’re living through, insurers will need to continue to embrace AI/ML to evolve in a constantly changing world.
Ryohei Fujimaki ([email protected]) is the founder & CEO of dotData, a spin-off of NEC Corporation and the first company focused on delivering full-cycle data science automation for the enterprise. Dr. Fujimaki is a world-renowned data scientist and was the youngest research fellow appointed in the 119-year history of NEC.