Agentic AI for Actuaries
A practical guide taking actuaries from AI fundamentals to building and governing autonomous actuarial systems — with working code and a running reinsurance case study.
About the Book
Agentic AI for Actuaries: From AI Foundations to Autonomous Actuarial Systems by Satya Sai Mudigonda and Dr Rohan Yashraj Gupta (FIA, FIAI) Published by ACTEX Learning
Artificial intelligence has moved past the chatbot era. The systems now entering insurance and reinsurance workflows don't just answer questions — they plan, use tools, retrieve data, check their own work, and carry out multi-step tasks with limited supervision. These are agentic systems, and they are arriving in actuarial work whether the profession is ready or not.
This book is written for actuaries who want to be ready — not as spectators, but as the people who design, challenge, and govern these systems.
Who it's for
The book is written for three audiences at once: aspiring actuaries working through their fellowship exams, practicing actuaries across pricing, reserving, life, health, pensions, and risk, and actuarial leaders — Chief Actuaries and CROs — who must decide where agentic AI belongs in their organisations and where it does not. Strong mathematical maturity is assumed; prior AI or Python experience is not.
What's inside
Across eighteen chapters in five parts, the book builds systematically:
- Part I — AI Foundations. How large language models actually work, explained without hype, with concrete actuarial examples leading every concept.
- Part II — Working with AI. Hands-on API usage, prompting, and structured outputs, introduced through realistic actuarial tasks.
- Part III — Agentic Building Blocks. Tools, memory, retrieval, and orchestration — the components that turn a language model into an agent.
- Part IV — Autonomous Actuarial Systems. Multi-agent architectures, reflection and self-correction patterns, and end-to-end actuarial workflows.
- Part V — Governance and the Road Ahead. Model risk management, validation, professional standards, and what agentic AI means for the actuarial career itself.
From Part III onward, every major concept is grounded in a single running case study: Meridian Re, a fictional composite reinsurer headquartered in Dublin with operations across Ireland, India, Singapore, and the US. Readers follow its Group Chief Actuary, reserving and pricing teams, CRO, and model risk function as they build, break, and govern real agentic systems — so the technology is never divorced from the professional context in which it must operate.
All code is real and runnable. A companion repository accompanies Chapters 9 through 17, with synthetic datasets and per-chapter documentation, built on a modern open agent framework so readers can reproduce every system in the book on their own machines.
Why this book is different
Actuaries are trained sceptics, and this book treats them that way. There is no AI evangelism and no vague reassurance. Claims are backed by named techniques, working implementations, and explicit discussion of failure modes, validation, and control. Where an agentic approach is the wrong tool, the book says so.
The result is a practical bridge between two professions' bodies of knowledge: rigorous enough for the actuary who wants to understand what is really happening inside these systems, and grounded enough for the leader who has to sign off on them.
Agentic AI for Actuaries is published by ACTEX Learning and available now.