AI, Risk Management, and Captive Insurance
December 09, 2019
My coeditor, knowing my affinity for technology, passed along a link to the new white paper from the Federation of European Risk Management Associations (FERMA) on artificial intelligence (AI). The paper, titled Artificial Intelligence Applied to Risk Management, was produced to answer the following strategic questions organizations are grappling with.
- How and why should our company be using and applying AI?
- What new liability or cyber challenges arise?
- What are the challenges for our workforce?
After reading the report, I thought it would be beneficial to integrate some of the ideas presented by FERMA on AI and risk management with captive insurance.
FERMA begins the paper by defining "artificial intelligence," using the European Union's definition released in April 2019. It states in part the following.
Artificial Intelligence (AI) systems are software (and possibly also hardware) systems designed by humans that, given a complex goal, act in the physical or digital dimension by perceiving their environment through data acquisition, interpreting the collected structured or unstructured data, reasoning on the knowledge, or processing the information, derived from this data and deciding the best action(s) to take to achieve the given goal.
Given the European Union's definition, I also searched for a US definition. In doing so, I came across the following paper: Artificial Intelligence and National Security, released on November 21 of this year. While not necessarily germane to this article, it did acknowledge that while the secretary of defense was tasked with developing a definition for Congress, at this time no official US definition of AI exists.
Artificial intelligence has only been made possible within the last decade by two interconnected trends—the collection and accumulation of huge amounts of data and the increased computational power of computers. As a result, businesses have sought ways of harnessing this data to optimize performance and gain market advantages. We have previously written about the increased use of big data within the insurance industry and encouraged captives to at least be conversant with this trend.
The FERMA paper suggests organizations begin to incorporate AI into their enterprise risk management (ERM) framework. There is an excellent AI action guide for risk managers included within the monograph. We encourage captive owners to download a full copy of the paper and take advantage of this operational guide. Granted, captives will be at different stages of developing their own ERM framework, but the proliferation of AI within the industry warrants its inclusion at some point in the ERM process.
Finally, let's focus on the "Benefits and Opportunities for Risk Managers Applying AI " chapter of the FERMA report. For captives that are struggling with how to begin incorporating AI/big data within their operational procedures, FERMA lays out the following eight-step process.
1. Define or identify problems you want AI to solve. For captive insurers, this will require that both the board and management become conversant with the technology. Use this first step as a springboard to become more comfortable with the concept.
2. Start small and develop a proof of concept to determine the viability of the project. Captives might think about working with their outside actuary on this step. Actuaries are well versed in designing and using models for data interpretation. Most actuarial firms today are well on the way to building out AI models. Your independent audit firm is also another potential source of support as auditors are using AI for much of their initial financial information analysis to seek anomalies.
3. Data quality. This is an easy concept to grasp. The adage "garbage in, garbage out" is especially true when asking AI to determine the best action to achieve a stated goal.
4. Build or buy. "AI may be necessary for every organization, but not every organization will have the resources to implement it on their own," the FERMA report states. This is especially true for captives, given their smaller size and limited resources—both personnel and monetary. Therefore, captives should seek to determine the potential vendors available to procure AI from and begin the process of interviewing and ranking these vendors. Captive associations can provide an additional benefit in this regard by acting as a clearinghouse for potential vendors.
5. Where will the requisite data come from? FERMA suggests asking the following series of questions.
- What data will we use?
- Do we have the right data?
- Do we need to combine our data with external data?
- If using outside vendors, do they even need our data?
For captive insurers, it is likely you will not have sufficient data for even the preliminary small-scale project you select for your test case. This then presents the opportunity to determine how best to utilize the data you do possess and to begin to accumulate additional data that can be used in the future.
6. Housing the data. Do we build our own data repository, or do we use cloud services? Again, for captives, this is a relatively easy question. With the proliferation of low-cost cloud services today, there is no need to build a big data warehouse and be responsible for its maintenance.
7. Bring your staff along. Implementing AI is a very big change within any captive. Since most employees are wary about change to begin with, it is critically important that management assure staff AI will help make their daily lives better and is not meant to supplant them in their roles.
8. Board and management buy-in. This adage is true of almost any change an organization decides to undertake. Staff will take their cue from senior management. If senior management is seen as wary or unsure of why AI is being used, staff will be even less willing to embrace the concept.
Big data and AI are here to stay. As noted previously, we believe there is a considerable amount of competitive advantage to be gained by early movers in this space. While captives may not be uniquely positioned to take advantage of AI immediately, those that begin seeking to understand the technology now will be ready to do so as soon as possible.
December 09, 2019