OpenAI engineer Charlie Guo contends that voice agents should not always speak back, urging developers to leverage speech-to-action interfaces that automate complex computer tasks, preserve conversational nuance through native audio processing, and enhance digital accessibility.