How to Build Voice Agents That Don't Fall Apart When You Interrupt - Jesse Hall, LiveKit
Voice agents fail in real-time conversation when interrupted, requiring human-like responsiveness to succeed
“Voice is the most natural interface we have.”
Jesse Hall from LiveKit presents on the core UX failure mode of real-time voice agents: they break down when users interrupt mid-conversation, causing customers to abandon to human operators. The talk frames latency as only part of the problem, arguing the real goal is human-like naturalness, with financial services as a high-stakes example where agent context-tracking failures carry real risk. This is an engineering practitioner talk on voice agent architecture, relevant to teams building or evaluating real-time voice AI products.