Overview
How might we design voice confirmation so people can trust and act on AI output?
When this pattern fits
- Essential for voice shopping, voice banking, smart-home automation, and any always-on agent where irreversible actions need an explicit spoken or multimodal confirmation before execution.
When to skip or lighten it
- Low-risk read-only queries.
- Confirm steps so long users disable voice entirely.
- Actions already protected by physical biometrics alone.
States
State model coming soon
Key UX elements
Key UX elements coming soon
Anti-patterns to avoid
Yes on any utterance after prompt with no detail repeat.
Confirm in voice only with no screen echo for purchases.
Same brief confirm for 1 dollar and 1000 dollar actions.
No cancel phrase documented.
How products use it
| Product | Implementation |
|---|---|
| Alexa | Voice PIN and confirm on shopping checkout. |
| Siri | Confirm transfers and sensitive HomeKit actions. |
| Google Assistant | Purchase confirmation on linked payments. |
| Apple Pay | Re-auth on device for voice-initiated pay. |
Implementation
Copy this prompt to generate a production-ready implementation in Cursor, Claude Code, Lovable, or any AI coding agent.
Generate a production-ready implementation of the "Voice Confirmation" AI interface design pattern.
Pattern Definition:Frequently asked questions
What must be repeated before confirm?
Action, amount, recipient, and irreversibility in plain language.
Voice confirmation vs voice commands?
Commands trigger actions. Confirmation is a gated second step for risky ones.
Multimodal confirm?
Phone tap or device button plus spoken yes for high value.
Cancel how?
No, cancel, or stop must abort without partial execution.