Physicians can learn a lot from talking to patients, and not just from the words they say, but from millions of data points from acoustic features of their speech, such as pitch, vocal cord vibration patterns, and micro-instabilities in the voice. Canary Speech has developed an AI-based diagnostic listening tool that can detect neurological and psychiatric conditions from physiological signals in the voice. We spoke to Kang Hsu, chief medical officer of Canary Speech, about how its diagnostic tool works, how it is embedded invisibly into clinical workflows, and how the company is driving physician adoption.