Billions flowing into voice AI

Investors are pouring billions of dollars into voice AI startups across model makers, customer service providers, meeting note-takers, and AI-powered dictation tools.

Full-duplex models exist but lack the moment

PolyAI CTO Shawn Wen says voice AI has reached full-duplex models that can speak while listening, but has not yet had its ChatGPT moment.

Speed and naturalness are the next hurdle

Wen says the next challenge is making reasoning very fast so models fetch answers quickly and conversations feel natural, not robotic.

“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.”

Building customer confidence matters

Wen says AI agents in customer service should not sound robotic and should give callers enough confidence they can solve problems.

Transcription accuracy breaks everything downstream

Otter CMO Alex Gay says Automatic Speech Recognition models miss important keywords, creating context problems. Poor transcription breaks all downstream actions and erodes trust.

“If your original transcription didn't have the accuracy that you needed, all the follow-up actions that you have become flawed.”

Digital twins need human emotion

Otter is working on digital twins to represent people in meetings. Gay says output voice must give the same emotive expressions as talking to a human.

“If you aren't able to have that with an avatar, then it's just a q and a chatbot.”

Transparency and trust are critical

Both PolyAI and Otter say voice AI tools must declare to customers they are being recorded or talking to AI. Transparency is essential for building trust.