An actuary who builds the environment.

Instead of speculating about what AI means for actuaries, I run experiments and publish the numbers, the failures, and the method. The agent does the case-by-case work; the actuary reviews, approves, and signs. Experience studies, predictive analytics, and interactive things you can poke.

NewOn Notes: The wiki that writes itself — Karpathy's LLM wiki, read as an actuary →
Featured · Asian Actuarial Conference 2026

Can an AI agent conduct an experience study?

Yes — when an actuary builds the harness. The agent held the pen; the actuary held the gate.

63 → 94 points
insight score /100, same submission
39/50 rules
drafted by the agent, all approved by the actuary
31M records
synthetic submission, 1M policies
Slides + the full experiment →

The other rooms


About

HyunSu Kim

HyunSu (HS) Kim, ASA — Senior Actuarial Consultant, Regional Life & Health Pricing, Munich Re Singapore; before that, Munich Re Korea and AIA Life. 15 years in life & health, on both the direct and reinsurance sides of the same data. I led experience studies and assumption setting for seven years, teach predictive analytics and AI at the Institute of Actuaries of Korea, and now work on integrating AI and advanced analytics into actuarial methodologies across APAC & MEA. The question underneath all of it: what should the agent do, and what must stay with the actuary? My answer so far: the agent does the case-by-case work; the actuary reviews, approves, and signs.

Everything on this site is personal work: built on my own setup, on synthetic data, with views that are my own.

LinkedIn · [email protected]