These execs think voice AI hasn’t reached its ChatGPT moment yet

Despite billions of dollars in investment fueling a wave of new startups, key executives in the voice AI space argue the technology has not yet had its breakthrough “ChatGPT moment.” At the recent HumanX conference, leaders from enterprise platform PolyAI and meeting assistant Otter agreed that while AI voices sound more human than ever, fundamental challenges in reasoning, recognition, and trust are preventing truly natural, reliable interaction.

Core Technical Hurdles Remain

Shawn Wen, CTO of PolyAI, noted that while the industry has achieved milestones like creating full-duplex models that can speak and listen simultaneously, the next barrier is speed. “The next challenge is to make reasoning very fast, so that the models can fetch answers quickly and the conversation feels natural,” Wen explained. He and Otter CMO Alex Gay both emphasized that the foundational layer, Automatic Speech Recognition (ASR), remains a critical point of failure. If the ASR misses keywords or misinterprets speech, all subsequent actions become flawed. For Otter, this means incorrect transcriptions and summaries; for PolyAI, it means failed customer service interactions.

• Reasoning Speed: Current models are often too slow to reason and respond, creating unnatural pauses in conversation.
• ASR Accuracy: Flaws in basic speech-to-text transcription undermine all downstream tasks and erode user confidence.
• Emotive Expression: For applications like digital meeting avatars, voices must convey human-like emotion to facilitate genuine strategic debate, not just Q&A.

The High Stakes of User Trust

Both executives stressed that overcoming these technical issues is directly tied to building user trust. “If your original transcription didn’t have the accuracy that you needed, all the follow-up actions that you have become flawed... You lose trust in the platform,” said Gay. For customer service agents, Wen noted that a caller’s confidence is built in the first few turns of a conversation. If the AI agent sounds competent and can solve the problem, users will trust it. Both also agreed on the need for transparency, making it clear to users when they are speaking to an AI or being recorded, viewing it as a prerequisite for building a stable user base.

The voice AI industry's path to mainstream adoption is not a matter of better vocal synthesis, but a race to solve core challenges in high-speed reasoning and foundational speech recognition accuracy. Without these, even the most human-sounding models will fail to build the user trust necessary for widespread commercial success.