The idea that voice will become the next major user interface has gathered momentum: investors have poured billions of dollars into startups focused on voice AI, spanning model developers, enterprise customer‑service providers, meeting note‑takers and AI‑powered dictation tools. Despite frequent announcements of new models and tools claiming humanlike speech and conversation, industry leaders say the technology has not yet reached a ‘‘ChatGPT moment’’—a clear, widespread inflection point of usability and trust.
Full‑duplex exists, but reasoning must be faster
Shawn Wen, chief technology officer of enterprise voice AI platform PolyAI, told attendees at the HumanX conference last month that his team has developed full‑duplex models that can speak while listening. However, he stressed the next challenge is speed of reasoning: models must fetch answers quickly so exchanges feel natural. “We have reached the milestone of developing full‑duplex models. The next challenge is to make reasoning very fast, so that the models can fetch answers quickly and the conversation feels natural,” Wen said on stage. He also argued that AI agents used in customer service must not sound robotic and need to instill enough confidence in callers that their problems will be solved.
Building user trust through repeated interactions
Wen described a path to adoption driven by trust: when voice quality is good enough and customers engage for two or three turns, they begin to build confidence. Over time, users may feel they don’t need to talk to a human if the agent can resolve their issue—an important step toward broader voice AI adoption.
Otter’s priorities: speaker ID, intent capture and digital twins
Alex Gay, chief marketing officer of meeting‑notetaker Otter, said speaker identification, intent capture and converting those signals into notes tied to organizational knowledge are key steps toward automation. Otter is also exploring digital twins that could represent people in meetings. For such technology, Gay said it’s essential that the output voice convey the same emotive cues as speaking to a human in a meeting context. “If you think about the meetings that you’re in right now, the best conversations that you have are where you can have debate, and strategic discussions, and when you feel like there’s a relationship that underpins it. If you aren’t able to have that with an avatar, then it’s just a q and a chatbot,” Gay noted.
Limits of ASR and the consequences for downstream automation
Although voice models have improved, AI assistants still often fail to understand users fully, and note‑taking tools sometimes produce incorrect transcripts or summaries. Wen pointed to automatic speech recognition (ASR) models frequently missing important keywords, which undermines capturing the full context. Gay agreed and emphasized that transcription was never Otter’s end goal but the layer on which productivity gains are built. If the original transcript lacks sufficient accuracy, all subsequent automated actions become flawed. He warned that when automation acts on incorrect transcripts, users lose trust in the platform. That makes continued improvement of ASR models critical because of the significant downstream impacts.
Transparency: notifying participants and disclosing AI participation
New voice tools also raise transparency questions. Products should make clear to participants when they are being recorded or talking to an AI. Otter said it aims to build trust among meeting participants and even in meetings without a bot present it considers approaches such as notifying everyone in chat when a meeting is being recorded. Wen from PolyAI similarly stressed the importance of making clear in enterprise calls when people are interacting with an AI.
Conclusion
Progress such as full‑duplex speech shows that voice AI is advancing, but industry leaders say the field still needs faster reasoning, more accurate ASR, emotionally faithful speech outputs and stronger transparency measures before voice systems achieve the broad, trust‑based adoption that would equal a ‘‘ChatGPT moment.’”
