Voice AI Where One Wrong Question Breaks the Rules - Shiv Indap, Upstart & Brooke Hopkins, Coval
Voice AI in regulated lending requires compliance guardrails where one wrong question triggers legal liability
“if something goes wrong, it's not just a customer experience issue. This could be a potential regulatory issue, which is much worse.”
Upstart runs thousands of daily voice agent calls for loan verification in a heavily regulated lending environment, where asking even a single prohibited question—like inquiring about the amount of medical-purpose funds—creates regulatory exposure beyond mere UX failure. Coval, founded by a former Waymo evaluation infrastructure lead, draws a direct analogy between autonomous vehicle reliability testing and voice agent assessment to address this. The talk highlights that enterprise voice AI deployment in regulated verticals demands evaluation infrastructure as rigorous as safety-critical systems, making compliance a first-class engineering concern.