Voice UX in practice: Handling ambiguity, interruptions, and failure

The concept of voice ordering is simple at its core: someone asks for what they want, and the systems (often unseen) transcribe the words, parse the intent, and confirm the order. But what happens in theory and what happens in practice are two very different things.

Should the system proactively suggest upsells, or wait until asked? When the user says “regular coffee,” should the system apply defaults or ask clarifying questions? How should it handle, “Actually, make that oat milk,” without forcing someone to start over?

Drawing on his experience building a voice-first ordering system for drive-thru and kiosk experiences, Kyle Alexander will share frameworks for addressing the tension around visual/voice coordination, fallback scenarios, and interruption handling. Attendees will leave with a fresh perspective and a handy set of tools for the future interface.

In this session, you’ll learn how to:

• Determine when voice systems should be proactive, deferential, or a combination of the two.
• Understand the coordination of sound and visual cues to build and protect user trust (especially when modalities present conflicting information).
• Design interruption handling that feels natural and organic.
• Identify and improve failure modes in multi-modal voice interfaces.
• Build conversation design systems that scale.