While climate change and cyber-threats are keeping risk management experts busy, they are also aware of governance issues and the specific flip-side of hyper-personalisation for insurance companies. All this, when they are ready to embrace big data, AI, and models to make their jobs faster and easier. Andrew D. Rallis, Global Chief Actuary, MetLife, and the immediate past president of the Society of Actuaries, gives some back-of-the-envelope calculations on the formula between technology and risk management in this interview. He also explains why it’s time for not just IQ or EQ but also AQ. Excerpts:
DQ: What has changed fundamentally about the actuary space in the pandemic phase?
Andrew D. Rallis: Actuaries are the leading professionals in finding ways to manage risk. The pandemic represents a type of risk that has not been encountered for several generations. It has had an impact across all parts of a company’s financial statements and daily operations, making it crucially important for the profession to use its risk management techniques. It has also led to public health issues, which makes it even more imperative for companies to evaluate risks, especially those associated with finance, insurance, and related fields as the frequency and size of claims can spike dramatically. So actuaries have been assessing and responding with new ways to mitigate the health, mortality, and financial risks posed by this new form of risk, particularly with respect to the financial security systems and programs that many actuaries are responsible for managing.
The risks and their interdependencies have become more complex. However, the tools and models that actuaries use have also become more sophisticated.
With accelerated automation, applying a level of emotional intelligence and providing insights will be of utmost importance which a machine is incapable of.
DQ: Has the job of insurance and risk analysis become easier or tougher now? Are businesses more open to the need for having the right models on their side? Are climate risks and ransomware getting up on the priority list?
Andrew D. Rallis: Insurance and risk analysis have become both easier and more difficult. The risks and their interdependencies have become more complex. However, the tools and models that actuaries use have also become more sophisticated. We now have the advantage of ‘big data and artificial intelligence that were not readily available just a decade ago. The opposing forces of risks and risk management are both developing rapidly. Insurance business leaders are aware of these rapid developments and are open to using new models, although most are also cognizant of the need for appropriate governance practices related to model development. Some of the biggest challenges of our time are tackling climate change and cyber threats. In this new environment, actuaries and their stakeholders are increasingly involved in assessing and responding to these global risks other than the pandemic.
DQ: What kinds of skills and tools have emerged as hot ones – for the actuary space – in the last two years?
Andrew D. Rallis: Actuaries already use advanced software to develop new methods of calculating and analysing data. Today, due to an increasing need to respond rapidly, actuaries no longer have the luxury of longer periods of time for risk analysis and planning. They have always been required to have strong technical skills and knowledge (IQ), and those who would rise to leadership roles also needed strong emotional intelligence quotient (EQ) skills. But the rapidly transforming environment of the past two years also requires adaptability quotient (AQ) skills. With accelerated automation, applying a level of emotional intelligence and providing insights will be of utmost importance which a machine is incapable of.
Tools that have been of importance include the ability to manage and appropriately utilise large quantities of data – much larger and more complex than have ever been available before – and build models that properly assess key components of the data and recognise interdependencies.
DQ: Have algorithms got better in forecasting and precision? What gaps – like poor or inadequate data, fraud, the wrong binning approach, modeling errors – confront them?
Andrew D. Rallis: More and more actuaries are using cross-disciplinary team approaches for modeling and forecasting. Healthcare professionals, social scientists, data analysts, and others work in conjunction with actuaries to improve their techniques through understanding the drivers of behavior and the causes of outcomes. So, while the algorithms have improved, those improvements come not only from advances in technology (which are certainly a significant factor) but improvements to the process of understanding the nature of risk and the ways to model and mitigate it.
Underwriting and risk analysis have been some of the earliest and strongest users and beneficiaries of AI and analytics in the insurance industry.
Investments in the foundation of data are becoming increasingly critical. Insurance firms have been increasing their focus on data transformation initiatives.
The gaps that exist are often created by faulty assumptions and understandings. For example, actuaries must avoid the common misunderstanding that correlation between two factors implies that causation exists. Often, the causation can be accounted for by a third or fourth factor, with there being no causal relationship between the original two correlated factors. Actuaries also know the importance of the axiom that “all models are wrong; some models are useful” – which is a way of indicating that we must remember that models are a tool for decision-making, but they are not a substitute for judgment. Actuaries bring to the table the ability to analyse the volumes of data that drive decisions but also the ability to put the decisions in the right business context.
DQ: What are your views on decentralisation and hyper-personalisation? Would these forces become staple factors soon?
Andrew D. Rallis: In some ways, hyper-personalisation and decentralisation are positive forces in the insurance world. They can personalise and enhance the client experience, making it feel more tailored and custom-made. Some of these approaches can also drive positive behaviors, such as better driving habits for monitored drivers or healthier exercise habits through fitness trackers for health plan participants. However, there is a limit to the degree of personalisation that can be supported within the insurance industry, for at least two reasons. First, insurers (and others) must be careful not to unduly discriminate in their business practices, and a high degree of personalisation can inadvertently lead to discrimination against members of certain demographic groups. Second, the reason that insurance can be provided by insurance companies is due to what’s called the “Law of Large Numbers” – that is, while outcomes for any single insured are unpredictable, the outcomes for a large group are on average highly predictable. Hyper-personalisation can break down the sizes of the insured groups to so small segments that the outcomes are no longer highly predictable, and the insurer is faced with either raising prices to account for the uncertainty risk or withdrawing coverage for groups for which the numbers are too small to be credible.
DQ: How much can AI and analytics help in making underwriting and risk analysis stronger?
Andrew D. Rallis: Underwriting and risk analysis (in addition to marketing) have been some of the earliest and strongest users and beneficiaries of AI and analytics in the insurance industry. Today, investments in building the foundation of data are becoming increasingly critical. Insurance companies have been increasing their focus on data transformation initiatives. This evolving data ecosystem has created an opportunity to drive innovation in AI and machine learning (ML) across many operational processes of insurance companies. As AI is further integrated into the insurance industry, the wait times for clients to have their policies approved and issued have often decreased dramatically, and the claims outcomes for the insurers have often improved significantly. These techniques have also been instrumental in detecting fraud in the underwriting and claims adjudication processes.