Moody’s Talks: Risk Reframed

Lessons from an AI engineer

Episode Summary

AI is changing how content is created, software is built, and risk management decisions are supported. But easier creation does not automatically mean better outcomes. The real challenge is preserving human judgment while grounding AI in trusted data and rigorous evaluation. In this episode, Han Lee, Senior Director of Data and AI at Moody’s, joins host Alex Pillow. Together they cut through the AI hype to explore what makes AI genuinely useful, trustworthy, and practical in risk management, from authentic writing and AI-enabled engineering to evaluation, trusted data, and total cost of ownership.

Episode Notes

AI is changing how content is created, software is built, and risk management decisions are supported. But easier creation does not automatically mean better outcomes. The real challenge is preserving human judgment while grounding AI in trusted data and rigorous evaluation.

In this episode, Han Lee, Senior Director of Data and AI at Moody’s, joins host Alex Pillow. Together they cut through the AI hype to explore what makes AI genuinely useful, trustworthy, and practical in risk management, from authentic writing and AI-enabled engineering to evaluation, trusted data, and total cost of ownership.

Key topics include:

• Why AI-generated content can converge toward a similar voice, and the importance of preserving original thinking and human expression
• How AI coding agents are shifting engineering bottlenecks from writing code toward deciding, planning, reviewing, and coordinating work
• Why domain expertise, trusted data, evaluation, monitoring, and auditability matter for AI used in risk and compliance
• How to think about build versus buy through total cost of ownership, including ongoing software, data, and operational maintenance

Additional resources:

• Han's new book: LLM Evaluation and Alignment, The Foundational Ideas
• Visit moodys.com to learn more