A common misconception about large language models (LLMs) is that they’re merely sophisticated chatbots that struggle with mathematical tasks. This view arises from seeing LLMs in isolation, as they are designed to predict the next token in a sequence rather than perform arithmetic. In other words, when an LLM is asked, “What is 56,372 x 34,252?” it sees only a string of tokens, not a math problem as a human user would. However, when equipped with the right tools (like a calculator), LLMs can tackle complex mathematical challenges, even solving problems from the advanced competitions like American Invitational Mathematics Examination (AIME).
This is where AI agents come in. AI agents are LLMs enhanced with the ability to use specialized tools, maintain context over time, and make autonomous decisions. Unlike a basic chatbot, these agents can break down complex problems into steps, choose appropriate tools for each step, and execute them sequentially with the broader objective in mind. Think of it as the difference between having a conversation with someone versus having a skilled assistant who can access calculators, code, and databases to help solve problems. As Microsoft notes, “AI agents are not only a way to get more value for people but are going to be a paradigm shift in terms of how work gets done.”
What this means for actuaries:
The path forward lies in learning to work alongside AI agents, much like how we mentor and guide interns. This collaboration requires two key competencies. First, we need to develop technical expertise in building with these agents. This includes understanding how to effectively communicate tasks, recognize agents’ capabilities, and acknowledge agents’ limitations. Second, and perhaps more crucially, we must learn to manage our customized agents. This involves identifying which tasks to delegate, providing guardrails to prevent errors, establishing clear success metrics, and creating feedback loops that enable agents to reason, learn and self-correct. The future of actuarial work is not about AI replacing actuaries but about AI amplifying actuarial expertise.
Sources:
• https://news.microsoft.com/source/features/ai/ai-agents-what-they-are-and-how-theyll-change-the-way-we-work/
• https://deloitte.wsj.com/cio/how-ai-agents-are-reshaping-the-future-of-work-127e8c6c
• https://www.anthropic.com/engineering/building-effective-agents
