Charlie Guo challenges the assumption that a voice agent must always answer with speech. Charlie Guo distinguishes speech-to-speech conversation, speech-to-action control, and event-to-speech notifications, arguing that these modes can be combined within one product.
Charlie Guo describes spoken form filling, creative workflows and computer control as opportunities to connect voice input to existing software actions. Visual state changes, notifications and other established interface patterns can communicate results without a spoken reply. Exposing application functions as model tools still requires guardrails and checks on safe, correct execution.
Charlie Guo discusses hands-free experiences, selective proactive speech and accessibility for users who cannot easily use a keyboard. Native audio can retain conversational cues lost in transcription, but the design question remains what the system perceives, which tools it may use, and when audio is actually the right way to communicate.
Watch on YouTube
