Voice AI Execs: ChatGPT Moment Hasn't Arrived — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- PolyAI CTO Shawn Wen said at the HumanX conference that despite full-duplex model releases, voice AI has not had its "ChatGPT moment," identifying fast reasoning as the next challenge so conversations feel natural.
- Wen said customer-service agents earn trust over the first two or three turns, after which callers accept AI if it solves problems — meaning voice UX must clear an initial confidence threshold before automation sticks.
- Otter CMO Alex Gay said speaker identification, intent capture, and layering organizational knowledge on transcription are the key automation steps, with the company also building digital twins that require emotive voice output for real meeting dynamics.
- Gay warned that if the original ASR transcript lacks accuracy, every downstream action becomes flawed and users lose trust in the platform, making transcription fidelity the foundation rather than the endpoint.
- Wen added that ASR models often miss important keywords, breaking context capture — a gap both executives identified as critical to enterprise viability.
- PolyAI and Otter both said transparency is non-negotiable: Otter notifies meeting participants of recording even when the bot isn't present, while PolyAI ensures callers know they are talking to AI on enterprise calls.
Why it matters: Investors have committed billions to startups across the voice AI ecosystem, and Wen and Gay are saying the foundational work — sub-second reasoning and transcription accuracy that survives downstream — still hasn't shipped, meaning enterprise customer-service and meeting-automation deployments remain gated on problems the field hasn't yet solved at scale.
Ask SkimNews
