We are collaborating with Carbon Five to build an Alexa Skill for the Amazon Echo. We want to share what learned along the way about voice UI scriptwriting.
We named our Skill “Standup,” after the meetings it’s designed to enhance. Standup helps agile teams keep their daily meetings on track and productive. After constructing personas and scenarios which helped determine our goals, the next step was modeling user interaction. This phase of the design process usually involves low-fidelity sketches that illustrate visual interactions, but voice UI doesn’t have physical interfaces. Instead of sketches, we needed a script that could simulate how a user might talk to Alexa (imagine a short skit).
We returned to our script again and again during the design process to adapt it to our latest discoveries. Here’s what worked for us:
Tip #1: Talk it out.
Talk is cheap! And it’s fast too, and best of all, it’s the medium that the product works in. Two people who know how Alexa works can model a conversation with low prep time and few materials simply by having a chat.
In our early conversations, we focused on what the user might be thinking. Later on, we paid more attention to how word choice and phrasing affects results.
Tip #2: Keep it snappy.
Tech users are accustomed to looking at screens and absorbing information at their own pace. Voice UI is a different animal—Alexa decides what the user hears, and the order they hear it in. Since Alexa’s voice is the only interface, her dialogue should be both brief and information-rich to keep users engaged. Especially for a Skill like Standup, designed for group meetings, brevity is key. A meeting goes as fast as its slowest speaker—make sure it’s not Alexa!
As Phillip Hunter, Head of UX for Alexa Skills at Amazon remarked at the 2017 O’Reilly Design Conference, “Be as informative as possible, but only [offer] as much information as is needed, and no more.” Alexa’s pace can feel slow at times, so wordy responses are out. Using a prototyping tool like SaySpring can help time her speech and keep things moving.
Tip #3: Cut the chatter.
A balanced conversation has give and take, and that holds true for VUI design. Interactions between Alexa and the user need to balance user effort (listening and speaking) with Alexa’s responses, while also minimizing both. Especially with common and frequent tasks, the back-and-forth should be as tight as possible.
We were ruthless in cutting unnecessary exchanges. An early script began the meeting with questions to help set an agenda, which we thought was a nice touch, but we found that it wasn’t worth the user effort. We decided to tuck this feature away in a separate command, allowing users to bypass it when it wasn’t necessary. This made a huge difference towards getting meetings up and running quickly.
Tip #4: Delight, don’t disrupt.
We were tempted at times to sacrifice efficiency for personality with unnecessary content. Personality is a big value-add for VUI, but not if it costs you function. If you want to to add humor, you could include a joke or a snarky jab, or you could get clever and look for moments where levity can shine through a thoughtful choice of words instead.
A great example: the Google Home trivia game. Google found a way to layer sassy personality over a tight script by assigning its players goofy names. Since players need to be identified anyway, it doesn’t take any extra time, but it’s also a crowd-pleaser.
Tell us about your scripts!
Are you designing a voice UI? We’d love to hear about your experiences in scriptwriting. Join our voice UI channel on the Cooper Friends Slack, send us your favorite articles on Twitter or send a note to email@example.com to chat about your voice UI strategy.