PolyAI’s Shawn Wen and Otter’s Alex Gay identified reasoning speed, transcription accuracy, and transparency as obstacles to more dependable voice AI.
PolyAI CTO Shawn Wen and Otter CMO Alex Gay described several remaining hurdles for voice AI in remarks reported by TechCrunch on October 11, 2026. Their comments point to a gap between systems that can sustain spoken exchanges and assistants that respond quickly, understand context, and reliably complete follow-up tasks.
Wen said full-duplex models—which can listen and speak at the same time—have reached an important technical milestone, but he sees fast reasoning as the next challenge. In customer service, he argued, an agent’s voice needs to be convincing enough to keep callers engaged while the system demonstrates that it can actually resolve their issues.
Gay highlighted a different set of capabilities for meeting assistants: identifying speakers, capturing intent, and connecting meeting information with an organization’s knowledge. Otter is also working on digital representations of people for meetings, he said, where conveying the emotional qualities of human conversation is important to making the interaction feel useful beyond a basic question-and-answer exchange.
Both executives pointed to transcription and context as practical weak spots. Wen said speech-recognition systems can miss key words, making it harder to interpret the full exchange. Gay said inaccurate transcripts can undermine later outputs and actions, such as summaries or productivity workflows, and weaken users’ trust when an assistant acts on faulty information.
The discussion also raised disclosure as a trust issue. Gay said Otter is considering ways to notify meeting participants when a meeting is being recorded, including situations where a bot is not visibly present. Wen stressed that people should know when they are speaking with an AI in enterprise calls. The executives’ remarks frame progress in voice AI not simply as more natural-sounding speech, but as a combination of speed, accurate understanding, dependable actions, and clear communication about AI involvement.