Recommendations for the use of AI in Actuarial Applications

In this insightful webinar, Dr. Małgorzata Śmietanka explores the transformative potential of Large Language Models (LLMs) in the insurance claims process. From automating document OCR and anonymizing sensitive data to reasoning over unstructured text, LLMs offer powerful capabilities to streamline and enhance key claims workflows.

The same processors that once fueled the Bitcoin gold rush and now train cutting-edge AI like ChatGPT are quietly revolutionizing actuarial science. While cryptocurrencies harnessed GPU power to mine digital coins, and AI uses it to process billions of data points, actuaries are discovering these chips can run complex mortality simulations in minutes instead of days. What began as hardware for gaming graphics has become our most powerful tool - turning overnight batch processes into real-time analytics and making trillion-scenario modeling suddenly practical. This presentation will explore how this transformative technology is reshaping our field, and why the actuaries who master GPU computing will lead the next wave of insurance innovation.

This research discusses the design, implementation, and performance of AI agents to meet regulatory standards and industry requirements. AI-enhanced ALM can better manage risks and optimize portfolio performance under various market conditions compared to traditional methods. The integration of AI in ALM processes presents promising opportunities for actuaries to advance their risk management frameworks, providing enhanced value to stakeholders in volatile economic landscapes.

Actuaries are busily trying to catch up with developments in Data Science and Artificial Intelligence. We cannot escape the feeling that we as a profession were caught by surprise. We have been a bit late, hopefully not fatally yet. The next revolution is waiting just around the corner. This comes in the form of Quantum Computing (QC). Will actuaries be ready to make use of it from day one? Full-scale adoption of QC still faces several technical challenges. There is however strong belief that QC will be reality in less than ten years. Today we mostly encounter QC with threats to today’s practices in cryptography. QC will make most of today’s practices insufficient (although there are some initiatives that are thought to be QC-proof). There will however be vast benefits from QC outside cryptography. QC will make it possible to do things that with today’s tools would be impossible (often solving problems in the so-called NP class). Or, they would be impossible in any realistic timescale even with computers get

As the regulatory landscape evolves to address nature-related financial risks, AI-based solutions are becoming increasingly vital for enhancing risk assessment. This study introduces AI-driven approaches that go beyond the Taskforce on Climate-related Financial Disclosures (TCFD) to encompass the broader Taskforce on Nature-related Financial Disclosures (TNFD). The TNFD framework covers physical, transition, and systemic nature-related risks, which are growing in significance for businesses. The study proposes three AI models: flood risk assessment using satellite imagery and Random Forest classification, biodiversity risk monitoring through neural networks applied to camera trap data, and systemic risk analysis based on public sentiment after natural disasters using Natural Language Processing (NLP) techniques like topic modeling. These AI tools offer innovative solutions for identifying, monitoring, and managing nature-related risks, aligning with emerging regulatory requirements beyond climate risk. By integrating these advanced technologies, the study bridges the gap between theoretical frameworks and practical applications, offering a comprehensive approach to proactive risk management in nature-related financial disclosures.

Machine learning models can enhance accuracy in actuarial functions like pricing and risk assessment. However, the Taiwanese insurance sector lacks experience in model fairness evaluation, which is required by local regulations. This study aims to evaluate the impact of applying fairness frameworks on the actuarial applicability of these models. This study applies a fairness framework to gradient-boosted tree models used to predict mortality, hospitalization, and surgery conditions of policyholders over an eight-year period. The framework consists of two stages: evaluation and improvement. In the evaluation stage, metrics like demographic parity, equal opportunity, and false positive rate parity are used to assess models’ prediction. In the improvement stage, the models undergo a three-step refinement process. Firstly, the models’ feature correlations are reduced during preprocessing. Secondly, fairness constraints are applied during model training. Lastly, models’ decision thresholds are adjusted to meet aforementioned metrics. To evaluate the impact of the fairness framework, the study compares the fairness-adjusted models to the original versions using both empirical and business metrics. Firstly, multiple empirical metrics like AUC, accuracy, and RMSE are used to compare the models’ performance. Then, the predicted outcomes of both models are used to classify policyholders into different groups, and the policy sale performance of each corresponding group is used as business metrics for evaluations. Our finding suggests that applying a fairness framework on models to comply with the Taiwanese actuarial regulations does not compromise their actuarial applicability. This study provides insights on implementing a fairness framework in the Taiwanese insurance sector, aiding the development of fairer and more reliable models for actuarial functions.