ClueCon Weekly with Hari Kishan: Voice AI Guardrails at Wire Speed

This week on ClueCon Weekly, host Jon Gray sits down with Hari Kishan, Director of Cloud Engineering at Manulife and author of the upcoming book Thinking Voice, to talk about what it actually takes to deploy voice AI safely at enterprise scale.

Hari makes the case that the model is just one component in a much larger system, and that most enterprise voice AI failures happen because teams treat the model as the whole system. He walks through how to build governance frameworks that validate and verify model output before it ever reaches a customer, why the best approach to PII is never sharing it with the model in the first place, and how to architect for both speed and safety simultaneously, rather than bolting on guardrails as an afterthought.

He also covers data classification strategies (red, orange, and green data tiers), the $25 million deepfake fraud case that made the rounds, voice biometrics, multilingual bot challenges, and why speech-to-speech is a whole new ballgame compared to traditional ASR.