Voice design is evolving as a critical discipline in the enterprise technology stack, particularly with the rise of voice AI. The challenge has shifted from technological limitations to designing effective voice user interfaces (VUI). Unlike designing for visual interfaces, voice design must account for the nuances of human conversation, including sensitivity to latency, natural language understanding, and the absence of visual cues.
The Salesforce guide outlines six principles of voice design, emphasizing the importance of naturalistic real-time voice experiences. These principles are grounded in conversation analysis and human interactional-linguistic behavior. Each principle is accompanied by audible examples and Agent Script encodings to illustrate improvements in voice agent performance.
The guide highlights three layers of voice design in Agentforce: prose prompt instructions, deterministic logic & variables, and voice channel configuration. These layers ensure that voice agents adhere to conversational rules, manage turn-taking effectively, and handle timing and sound appropriately. The selection of a voice is also crucial, as it must align with the agent's persona and meet the needs of callers.
Successful VUI design requires respecting the systematics of conversation and the limits of human cognitive processing. The six principles provided serve as a foundation for creating voice agents that deliver seamless and natural interactions, ultimately enhancing user experience and business value.