Your Financial Agent Needs to Show Its Work - Anirudh Veeraragavan, Fieldguide

Why watch

Anirudh Veeraragavan explains why financial AI agents must be auditable and defensible, walking through four critical controls: authorization to preserve human review history, execution records for step-by-step replay, evidence provenance tied to specific document versions, and eval data auditing to prevent misleading accuracy scores. The talk uses bank reconciliation audit as a concrete example and extends the pattern to credit decisions and anti-money-laundering. For engineers building AI systems in regulated financial industries.