In the meticulous world of actuarial science, where statistical rigor and business insight converge to assess risks and forecast financial outcomes, a quiet evolution is underway. This evolution is driven by the nuanced capabilities of generative artificial intelligence (Gen AI), a branch of AI that has gradually gained recognition within the industry.
What is generative artificial intelligence?
Gen AI is the emerging field of AI and was popularized in 2022 by OpenAI ChatGPT, a Gen AI chatbot based on Generative Pre-training Transformer architecture. It encompasses algorithms that can produce new content, whether it’s text, visuals, audio or even computer code. Gen AI includes a variety of methods, such as sophisticated text generators, deepfake technology and systems that convert text descriptions into images.
Among the various generative technologies, the key advancement pertains to the ability of machine-learning models to generate responses in natural language that closely resemble those produced by humans, often making them nearly indistinguishable.
For actuaries who grapple with vast amounts of unstructured data, Gen AI offers a promising avenue to transform data into actionable insights, enhancing the precision and efficiency of their work.
While generative AI encompasses a broad range of techniques and applications, large language models (LLMs) represent a significant breakthrough in the field.
What is a large language model?
LLMs are central to Gen AI. These models are part of two primary types of Gen AI/generative adversarial networks, which excel in creating images, texts and translations, and LLMs themselves. LLMs are specifically designed to understand the statistical relationships between words and phrases, enabling them to produce text that is both coherent and grammatically correct.
These models are trained on vast datasets, which makes them highly proficient in mimicking human communication. This capability provides actuaries and other professionals with powerful tools to efficiently manage and analyze complex data.
A critical aspect of how these models work involves tokenization, a process vital to text analysis and machine learning. Tokenization breaks down unstructured text data into smaller units, or tokens, assigning each a numerical index, which simplifies the way models understand and process language.
Navigating the divide: traditional machine learning vs. Gen AI
The AI landscape is characterized by traditional machine learning, with its discriminative models for classifying and predicting outcomes, and Gen AI, which fosters innovation by creating entirely new data. The distinction between the discriminative and generative models is vital. Discriminative models excel in classification and prediction tasks, making them suitable for analytical applications. In contrast, generative models are unparalleled in their ability to create and innovate, making them ideal for tasks requiring new content generation.
What are the commonly seen patterns for Gen AI solutions in financial services?
In the dynamic world of financial services, Gen AI is a transformative force, introducing advanced capabilities that redefine industry standards. Here’s how it’s making an impact:
Semantic search
Enhances management and searching capabilities within large datasets at financial institutions. This includes:
- Document management and organization: Analyzing document content, categorizing documents, generating descriptive file names and organizing documents based on their priority, content and relevance.
- Knowledge retrieval and question answering: Providing rapid responses to queries about policies, claim files, regulatory documents, etc.
Summarization
Gen AI tools quicky summarize complex financial reports and communications, helping professionals grasp essential points and make informed decisions without sifting through extensive data.
- Document and file summarization: Summarizing claim files, case histories, inspection reports, appraisal reports, medical documents, demand packages and regulatory documents.
- Conversation and communication summarization: Summarizing various forms of communication such as calls, emails and chat interactions.
New content generation
AI’s capacity to generate personalized content, from customer communications to financial advice, is revolutionizing customer engagement and advisory services in finance:
- Automated correspondence generation: Drafting follow-up emails, status-update letters, coverage-position letters, subrogation demand packets and contention letters.
- Analysis and recommendation generation: Analyzing files, policy details, past interactions and guidelines to generate recommendations for next steps, coverage determination and cross/up-sell opportunities.
Translation and code generation
These Gen AI features are reshaping cross-border financial operations and fintech development by facilitating language translation and converting natural language into code, thus removing barriers and speeding up financial application development. Use cases include:
- Technical content generation: Generating user stories, test cases and dummy data for testing purposes.
- Code generation: Producing code based on provided problem statements.
Impact of Gen AI on the financial-services sector
By adopting these Gen AI functionalities, the financial-services sector is addressing challenges associated with manual processes and data accessibility, enhancing efficiency and fostering innovation.
Executives across industry believe Gen AI will disrupt their industry, provide value to their business and have a positive impact on their workforce.
How can Gen AI change actuarial work?
Gen AI is set to transform actuaries’ work, traditionally a blend of statistics, business insight, and intuition for risk assessment and financial forecasting. This AI introduces tools that generate synthetic data for innovative uses, automating tasks like scenario analysis, model validation, underwriting, etc. It allows actuaries to consider broader outcomes and tailor insurance policies by analyzing personal risk factors in depth.
Additionally, Gen AI can draft content such as reports, saving time and ensuring consistency. It also supports actuaries in coding, even for those with limited skills, and offers personalized learning tools.
Overall, Gen AI is enhancing actuarial analysis, product innovation and operational efficiency, allowing actuaries to focus on strategy and advisory roles, thereby adding more value to their organizations and clients.
Actuarial applications: a new frontier
In the realm of actuarial science, Gen AI is revolutionizing traditional processes, bringing forth new efficiencies and capabilities. Here are some key use cases where Gen AI is making a significant impact.
| Use Case | Description | Benefits |
| Analysis report generation | Gen AI automates the creation of content and narrative summaries for reports, providing insights on technique-run results and financial metrics. It can draft sections of actuarial and financial memos in standardized formats. | Delivers insightful analysis efficiently and ensures consistent, high-quality documentation for actuarial reporting. |
| Actuarial model validation | A Gen AI knowledge bot simplifies the understanding of complex actuarial models and techniques, enabling quicker adoption. It provides concise explanations, simulates scenarios and offers interactive learning. Anomaly detection identifies inconsistencies early. This tool not only streamlines the validation process but also ensures compliance with the latest standards, enhancing decision-making and risk management with minimal effort. | Streamlines the validation process, reduces time and effort, ensures model reliability and compliance with standards, and enhances decision-making and risk management. |
| Automation of model executions and scenario analyses | AI chat agents automate model executions and scenario analyses, streamlining pre-model run processes like data aggregation and filtration, and initiating calculation engines for actuarial computations. Cross-referencing against various data sources is also facilitated. | Boosts efficiency and decision-making, frees up expert resources for higher-value tasks. |
| Actuarial model development | Gen AI assists in writing code from text prompts, reviewing and debugging code, suggesting improvements, adding comments for documentation, creating unit and regression tests and converting code between different programming languages. For example, it can convert Excel files to Python code to harmonize systems and improve computational performance. | Optimizes model building and running, enhances code production, accelerates development, simplifies knowledge transfer and ensures thorough documentation model management. |
| Smart meeting transcriptions and summaries | AI transcription tools can accurately record meetings, distinguish between different speakers, and generate summaries that highlight key discussion points and actions. | Saves time, captures detailed information and improves meeting productivity. |
| Personalized learning assistants | Gen AI enhances learning by offering customized content, interactive questioning, and automated feedback on quizzes and assignments. For example, it can be used with software like AXIS to create a personalized chatbot that helps new juniors learn the software. Additionally, a chatbot can interpret IFRS 17 standards, aiding new hires in understanding this complex regulation. | Improves learning efficiency, provides personalized educational experiences and speeds up knowledge acquisition. |
| Enhanced underwriting | Gen AI streamlines the underwriting process by checking databases for consistency, synthesizing data for unique coverages and providing relative risk scores. It also evaluates new groups for group coverages. | Enhances underwriting efficiency, improves risk assessment, reduces manual errors, and provides timely information for better decision-making. |
| Data management and analysis | Gen AI supports data enrichment, manipulation, validation, anomaly detection and automated analysis. | Efficiently generates large volumes of test data, addresses privacy concerns with synthetic datasets, enhances analytical techniques, saves time, reduces manual effort and ensures consistent data formatting. |
Embarking on the Gen AI journey
As the actuarial profession undergoes digital transformation, Gen AI stands out as a key innovation, offering actuaries the chance to significantly improve their analytical skills and efficiency. Actuaries must start by building a strong foundation in Gen AI principles, with online courses and workshops as ideal resources for gaining knowledge and practical experience.
Actuaries can explore Gen AI tools such as GPT for language processing or open-source models like Mistral or Llama. These tools can be seamlessly integrated into actuarial workflows through application programming interfaces. Initiating with low-risk projects provides a secure environment for actuaries to experiment with Gen AI and evaluate its efficacy compared to conventional methods.
As actuaries become more familiar with Gen AI, it can be seamlessly incorporated into daily tasks, automating routine work and enhancing data analysis. It is crucial to develop actuarial-specific best practices for Gen AI usage.
Actuaries should not merely be users of Gen AI but should play a central role in its implementation. With their deep understanding of data and insurance models, actuaries are uniquely positioned to ensure that Gen AI is used effectively and responsibly within the profession.
About the authors
Simon Girard, FCIA, is a partner in the Insurance and Actuarial Advisory Services practice of Ernst & Young (EY) and leads the practice for Canada. As a partner, he serves companies throughout North America and is committed to innovation and advancing the actuarial profession. He currently sponsors the development of advanced analytics solutions tailored toward actuarial and strategic applications for the life and annuity insurance industry. His other focus areas include accounting change (IFRS 17) and actuarial and finance transformation. Simon is also a Fellow of the Society of Actuaries and a chartered financial analyst charterholder.
Cem Unlubayrak is a senior manager in the actuarial practice at EY Canada. With over 10 years of experience in the insurance sector, he combines engineering and actuarial skills to drive transformative innovation for insurers. His focus areas include the implementation of IFRS 17, the development of high-value use cases involving artificial intelligence, and the analysis and management of emerging risks.
This article reflects the opinion of the authors and does not represent an official statement of the CIA.
