The latest voice AI developments are promising, but execs believe the technology hasn’t yet reached its potential.

Shawn Wen, CTO of PolyAI, says full-duplex models are a step forward, but the next challenge is making reasoning fast enough for natural conversations. Alex Gay from Otter emphasizes the importance of speaker identification and emotive expressions to build trust.

Despite improvements, ASR models still miss important keywords, leading to flawed transcription and loss of trust. Otter continues to work on accuracy, while PolyAI stresses the need for transparency in enterprise calls.

The future of voice AI may depend on making it sound more human, engaging, and reliable, but the journey is far from over.