Get Plugged In Practical AI for Actuarial Modeling
10 snips
Mar 4, 2026 
Igor Nikitin, CEO and co-founder of Nice Technologies and an actuary-software engineer-entrepreneur, discusses practical AI for actuarial modeling. He shares real-world AI uses for model development, review, and documentation. He warns about hype, data limits, black-box outputs, vendor claims, and scaling costs. He offers rollout tips: start small, set privacy and governance, require citations and human sign-off.
AI Snips
Chapters
Books
Transcript
Episode notes
AI As Reviewer And Development Assistant
- AI functions well as a reviewer and as an assistant for model development, speeding checks and drafting calculations.
- In PRT work Igor Nikitin uses AI to spot data issues across tens of thousands of pensioners and to draft tests, formulas, and efficiency improvements.
AI Success Depends On Public Training Data
- AI performance depends on available public training data; tasks with scarce public examples are unlikely to be reliable.
- Nikitin notes actuarial modeling often lacks large public datasets, so credibility issues mirror classical actuarial credibility problems.
Compare AI To Hiring For Real Costs
- Compare AI versus human alternatives and measure end-to-end cost and effort before choosing AI for a task.
- Igor ran an experiment: human developer coded pre-retirement death benefit in 6 hours while a prompt-only AI approach took 8 hours and required heavy review.
