Expanding the Actuarial Toolkit with Generative AI

Author: Klaas Stijnen, Co-CEO of Montoux

Introduction

In the ever-evolving technology landscape of the insurance industry, the emergence of generative artificial intelligence (gen AI) stands out for its transformative potential. This technology will enable insurers to re-think and modernize every aspect of the enterprise, from acquisition and management of customers to re-imagined back office operations.

Gen AI has put actuaries at a critical juncture where they will need to embrace change and adapt their skills and methodologies. Effective use of gen AI will enable actuaries to be vastly more productive and to generate deeper and more insightful analysis, in real or near real time. From automating

model coding and documentation to refining decision-making processes, gen AI equips actuaries with the capabilities to perform their tasks with greater efficiency, accuracy, and depth. As we venture into this new era, the actuarial profession must navigate the balance between technological innovation and the indispensable human judgment that has always been at its core.

“With generative AI, we observe for the first time that AI can not only have incremental, but disruptive influence on lots of processes and business models” Fabian Winter, Chief Data Officer - Munich Re Group.

The following article provides examples of how gen AI will augment and amplify the work of actuaries, now and into the future.

What is Generative AI?

Microsoft defines gen AI as: “A form of artificial intelligence in which models are trained to generate new original content based on natural language input. In other words, you can describe a desired output in normal everyday language, and the model can respond by creating appropriate text, image, or even code output”

Financial services organizations around the world are already transforming themselves with gen AI. Actuarial departments are under increasing pressure from their organizations to identify and trial the use of gen AI. Whilst most are very early in this journey, we see a great deal of interest and in several cases, investment in building out gen AI capabilities.

Embracing Gen AI in Actuarial Work

The practical application of gen AI in actuarial tasks is evolving rapidly, driven by the technology's ability to automate complex processes and generate insightful analyses. As highlighted by industry professionals, gen AI excels in specific, narrowly defined tasks, such as coding, generating test cases, and synthesizing information from vast datasets. However, its effectiveness in tasks requiring deep contextual understanding and multi-step reasoning, such as comparing actuarial model outputs, currently necessitates detailed, step-by-step instructions from actuaries.

This collaboration between actuaries and gen AI tools mirrors the dynamic between pilots and autopilots, emphasizing the importance of human oversight and judgment. The analogy of gen AI as a "4-year-old with a huge photographic memory and 20 university degrees" captures the essence of its capabilities and limitations, underscoring the need for actuaries to guide and refine its outputs.

Current Actuarial Gen AI Use Cases

At Montoux we are supporting insurers to enhance model documentation and governance using gen AI. This use case is an obvious starting point for gen AI adoption - it automates highly manual and unpopular work and delivers real business value through improved efficiency and risk management. Montoux Model Copilot, the first gen AI copilot built specifically for actuaries, reduces 80-90% of the highly manual and unpopular actuarial work that goes into model documentation, as well as enabling automation of previously labor intensive tasks such as model comparison. Both this use case and those mentioned below are reflective of the current state of gen AI implementations across major enterprises - they are focused on productivity and are primarily augmentation tools or assistants - i.e. the AI needs to work in conjunction with a human.

It is fair to say that the majority of actuarial departments are moving cautiously towards gen AI adoption, however those that are early adopters are already seeing value from the technology, particularly as a means to automate previously manual tasks. One of the ways in which gen AI differs from previous technological development leaps such as cloud or broader machine learning/deep learning is that the barrier to be able to evidence value from a gen AI investment is relatively low. Put simply, seeing is believing. As most of us have witnessed through use of tools like ChatGPT, Midjourney or Perplexity AI, it is possible to see value in gen AI almost immediately.

The gen AI use case that actuaries tell us they are implementing most commonly today is document analysis and report generation. Actuaries are using readily available tools like ChatGPT to ingest documents such as competitor statements and reports to assess and synthesize key points for further analysis. Similarly they are also using gen AI to accelerate the time-sensitive analysis that goes on during processes such as M&A or reinsurance contract negotiation. Actuaries who work in reporting roles are also telling us they are using gen AI to support the creation of a ‘first-draft’ report, which may require the ingestion and analysis of different data sources including model output and reports/documentation. The human actuary then works from the AI created draft, significantly improving their productivity and reducing reporting process time. The use of gen AI as a productivity enhancer in this way is also consistent with its use in other industries such as law and education.

Challenges

While the adoption of gen AI in actuarial functions promises numerous benefits, it is not without its challenges. These include ethical considerations, privacy concerns, regulatory compliance issues, and the readiness of technical infrastructure and skills.

Ethical Considerations and Privacy Concerns: The use of gen AI in managing personal data raises ethical questions, particularly around privacy. Actuaries must ensure that the use of gen AI does not inadvertently breach data privacy rules related to PII/PHI. Ensuring that gen AI systems are designed with privacy-preserving techniques and can be transparently assessed and audited is crucial in maintaining the trust of policyholders and the public.

Regulatory Compliance and Governance Challenges: The regulatory landscape for gen AI in insurance is still evolving. Actuaries face the challenge of navigating uncertain regulatory waters, ensuring that the use of gen AI complies with existing laws and anticipating future governance frameworks. This requires a proactive approach to regulatory engagement and compliance, with a keen eye on how AI applications in actuarial work align with standards for fairness, accountability, and transparency.

Technical Readiness and Skills Gap: For many insurance companies, the technical infrastructure may not yet be fully equipped to support advanced gen AI applications. Similarly, there may be a skills gap within actuarial teams, who traditionally may not have been trained in AI. Addressing these gaps requires investment in technology and training. Actuaries will need to upskill or reskill, embracing new tools and methodologies to leverage AI effectively. This transition is not just about acquiring new technical skills but also about fostering a culture of innovation and continuous AI learning within the organization.

Overcoming these challenges is essential for actuaries to harness the full potential of gen AI. It demands a collaborative effort, balancing innovation with ethical responsibility, regulatory compliance, and the development of necessary technical capabilities.

For actuarial departments looking to get started with gen AI, Montoux has developed a simple “How to Start Guide”.

The Future of Generative AI in Actuarial Work

Gen AI capability, and in particular the capabilities of large language models (LLMs) are evolving so rapidly that crystal gazing is a somewhat difficult challenge (there has been sufficient media coverage of the potential impact of general artificial intelligence to evidence this point!). Nevertheless, based on our conversations with customers there are a few key areas where we expect to see gen AI impact actuarial work in the coming 6-24 months.

AI-assisted Actuarial Model Development

Software engineering has been revolutionized through the use of AI coding-assistants such as Github Copilot and Amazon Codewhisperer. Actuarial model development is fundamentally the development of code, and so there is every reason to believe that it too will be transformed by gen AI. At Montoux we see this as a massive gain for actuaries, who are often acting in software engineering roles despite not being trained in that craft.

Real-Time Decision Support with Richer Insights

As the internal and external data available to gen AI tools increases in both volume and quality, and the capabilities of LLMs continues to evolve, we see a near-future whereby actuaries can get real-time answers to complex questions. The days of ‘we’ll come back to you in a few weeks after we’ve done the analysis’ may be capped, as the capabilities of gen AI enable more nimble and dynamic decision making. The results won’t be perfect, and may need validating by humans, but organizations that adopt an AI-first approach to responding to movements in markets and competitor behavior stand to gain a significant competitive advantage.

Increased Model Transparency and Wider Access to Actuarial Models

A large number of the world’s actuarial models are contained in legacy actuarial modeling systems that are either black-box or highly complex in nature. This results in a number of challenges, including vendor-lock in, internal and external key-person risk, and constrained ability to innovate in modeling. Outside of the actuarial domain there is significant progress being made in the use of gen AI to support the migration away from legacy code bases such as COBOL, and it is entirely likely that some of these techniques will creep into the actuarial modeling world. Furthermore, the ability to query a model in natural language (already possible using tools like Model Copilot), increase the transparency and understanding of actuarial models both within actuarial teams and across the enterprise. As well as significantly improving productivity, this also improves the utility of the model, opening it up for use cases other than Valuation.

Strategic Recommendations

To capitalize on the opportunities presented by gen AI, actuaries and insurance companies should consider several strategic actions:

- Investment in Specialized Training: Building AI literacy and technical proficiency across actuarial teams is essential. This can be achieved through targeted training programs, workshops, and continuous learning opportunities.

- Infrastructure Upgrade: Ensuring the technical infrastructure is robust enough to support AI applications, including secure data storage and processing capabilities.

- Collaborative Research Initiatives: Engaging in joint research projects with tech companies and academic institutions can provide access to cutting-edge AI developments and innovative practices.

institutions can provide access to cutting-edge AI developments and innovative practices.

- Ethical AI Frameworks: Developing and implementing guidelines for the ethical use of AI in actuarial work, emphasizing transparency, fairness, and privacy.

By embracing these future directions and strategies, actuaries and insurance companies can ensure they remain at the cutting edge of their profession, leveraging gen AI to drive efficiency, accuracy, and innovation in their practices.

Conclusion

The integration of gen AI into actuarial practices signifies a transformative shift in the insurance industry, offering opportunities for enhanced efficiency, precision, and innovation. By embracing this change, actuaries can expand their toolkit, adapt their skills, and continue to play a pivotal role in navigating the complexities of risk and uncertainty. Most actuaries we speak to want to do more with less, and this is exactly what gen AI enables!

Perhaps most importantly, actuaries shouldn’t feel threatened by the emergence of gen AI. At Montoux we use gen AI extensively as a means to augment and amplify the capabilities of our actuaries - not as a displacement for actuaries. Actuaries who can effectively use gen AI to augment and amplify themselves will become increasingly more valuable to the organizations they work for.

We encourage actuaries and insurance professionals to actively engage with gen AI technologies. Embrace innovation, invest in continuous learning, and participate in strategic planning to steer the future of actuarial work. Early adopters of gen AI can gain a significant competitive edge by optimizing processes, enhancing decision-making, and offering innovative products that meet evolving customer needs.

The journey toward gen AI adoption is complex and requires a concerted effort across the industry. However, by embracing these challenges and opportunities, actuaries can ensure a future where gen AI not only enhances their work but also elevates the entire insurance industry to new heights of success and service.

Klaas Stijnen is Co-CEO & Co-founder, Montoux. Klaas can be contacted at [email protected]