In this research paper, we demonstrate that a generative AI model can effectively empower and enable actuaries by assisting in producing a first draft of summary narratives to explain the analysis of movements in best estimate liability (BEL). The aim is not to fully automate the process but to prepare a first draft, thus empowering and enabling focus on other valuable work.

While initial development required significant time and effort, once established, the model enabled quick and efficient generation of reports for different periods, proving it could increase the speed and efficiency with which regular reports/narratives of results can be produced.

Key benefits and implications of this test generative AI model for insurers include:

• Efficiency and speed: The model can rapidly process all input data. Once initially set up, the model can be run in less than five minutes.
• Consistency and standardisation: The generative AI model can generate narratives using consistent language and structure, which minimises the variability and potential bias in reporting.
• Customisation and adaptability: The model can be fine-tuned to meet specific reporting requirements, narrative styles and regulatory formats.
• Enhanced insight generation: The model can highlight trends, anomalies, and drivers of change
• Scalability: The generative AI model can be easily scaled to produce narratives for multiple lines of business, product types or reporting periods