Voice AI Awaits Its ChatGPT Moment

Shawn Wen, CTO of enterprise platform PolyAI, believes the current focus on models that can speak while listening misses the mark on true utility. He argues that the next hurdle is latency; reasoning must become instantaneous for conversations to feel genuinely natural. Beyond technical speed, Wen emphasizes that AI agents in customer service must shed their robotic veneer to earn user trust. Once callers feel confident that an agent can resolve their issues without human intervention, he predicts a shift in how enterprise support operates.

Alex Gay, CMO of Otter, views the challenge through the lens of organizational intelligence. For meeting automation to succeed, systems must master speaker identification and intent capture while integrating deeper knowledge of company workflows. Otter is currently exploring digital twins capable of emotive expression, as Gay warns that without the ability to facilitate debate or maintain a human-like rapport, AI avatars remain little more than glorified Q&A chatbots.

Accuracy remains the primary barrier to adoption. Both Wen and Gay acknowledge that current Automatic Speech Recognition models often fail to grasp critical context, leading to flawed summaries and broken trust. Because downstream actions rely on the initial transcript, any error ripples through the entire productivity chain. As these tools proliferate, both companies stress that transparency is mandatory—users must be notified when they are interacting with an AI, a necessary step toward normalizing the technology in professional environments.