Jesse Hall of LiveKit argues that voice AI systems must prioritize humanlike behavior over raw speed. Developers should treat latency as a budget, measure what callers hear, and test complete conversational pipelines rather than isolated benchmark scores.