Voice AI Lacks ChatGPT Moment as 2 Key Gaps Slow Reasoning and Transcription
Updated
Updated · TechCrunch · Oct 11
Voice AI Lacks ChatGPT Moment as 2 Key Gaps Slow Reasoning and Transcription
1 articles · Updated · TechCrunch · Oct 11
Summary
PolyAI CTO Shawn Wen said voice AI still falls short of a breakout moment because models cannot reason fast enough to keep conversations natural, even after full-duplex systems learned to speak while listening.
Otter CMO Alex Gay said weak speaker identification, intent capture and transcription accuracy still undermine automation, because errors in the first ASR layer cascade into flawed summaries and actions.
Enterprise use cases raise a second hurdle: trust. Wen said customer-service agents must sound confident enough to handle problems without a human, while Gay said meeting avatars need humanlike emotional expression to support real debate.
Both companies also stressed transparency, saying users should be told when they are speaking to AI or being recorded as voice tools spread across call centers, note-taking and digital-twin meeting products.