Smart devices are everywhere — and interacting with them via voice is becoming common. Here are 4 guidelines for designing great voice-first apps.

I built a voice-first, Apple Watch Siri shortcut to help me reach my hydration goals — 8 glasses a day needs the support of good habits and reliable tracking to achieve. Through a series of iterations, I established these guidelines to designing great voice apps:

  1. Frequency: How often will the app be run? A user might call a water logging app ten or more times a day, but an Alexa news app just once at lunch. Conversation design needs to be planning accordingly.

  2. Avoid confirmations: at first, I thought verbal confirmation — “I think I heard you say 12, is that right?” — helps avoid bad data. I learned that Siri’s voice recognition was very accurate, and the confirmation became an irritating delay. Use data guard rails instead: validate the datatype, confirm it’s within an expected range, and throw an error if it’s outside expectations.

  3. Mix it up: use simple variations. Human ears are tuned to differences in conversations: small expression changes (good start, keep it up, great work, etc.) are emotively supportive and appreciated.

  4. Keep it tight: design conversations with concision. My first app iterations used lots of qualifiers and confirmations interrupters, like “OK!” and “Done” when they weren’t necessary. Users want a result, not a chat.