PolyAI and Otter Say Voice AI Still Lacks Its ChatGPT Moment
PolyAI and Otter executives say voice AI still lacks fast reasoning, accurate transcription, emotional nuance and clear disclosure needed to earn trust.
Summary
As of October 11, 2026, investors have poured billions of dollars into voice AI model makers, enterprise customer service, meeting note takers and AI dictation, with new releases arriving weekly. Yet PolyAI CTO Shawn Wen told the HumanX conference last month that voice AI has not reached its ChatGPT moment. Full-duplex models can listen while speaking, but reasoning and answer retrieval remain too slow for consistently natural conversations. Wen said customer service agents must sound less robotic, identify themselves as AI, solve problems and earn callers’ confidence within the first two or three turns.
Otter CMO Alex Gay said useful automation requires accurate speaker identification, intent capture and transcripts connected to organizational knowledge. Otter is developing digital twins to represent people in meetings, but their voices need humanlike emotion to support debate, strategy and relationships rather than basic question and answer exchanges. Automatic Speech Recognition models still miss important keywords, while inaccurate transcripts corrupt summaries and follow-up actions, undermining trust. Otter is improving transcription and considering chat notifications whenever meetings are recorded, even without its bot present, while PolyAI also supports clear AI disclosure during enterprise calls.
Positives
- Full-duplex models can now listen and speak simultaneously, a milestone PolyAI CTO Shawn Wen says voice AI has reached.
- Billions of dollars are funding voice AI model makers, customer service platforms, meeting note takers and AI dictation tools.
- Otter is connecting speaker identification, intent capture and organizational knowledge to enable automated meeting actions.
- Otter is developing digital twins intended to represent people in meetings with humanlike emotional expression.
- Otter and PolyAI support clearer notification when meetings are recorded or enterprise callers are speaking with AI.
Risks & concerns
- Reasoning and answer retrieval remain too slow for voice AI to achieve the natural interaction associated with a ChatGPT moment.
- Automatic Speech Recognition models can miss important keywords, preventing systems from capturing complete conversational context.
- Inaccurate transcripts can corrupt summaries and follow-up actions, causing users to lose trust in Otter’s platform.
- Meeting avatars without emotional expression could be limited to basic question and answer exchanges instead of debate and strategic discussion.
- Customer service agents may lose callers unless they sound natural and prove useful within the first two or three turns.