Automating Customer Service
Designed the "Draft from Thread" and feedback report features to help automate customer service emails.
Product Designer (Contract)
What is Promptly?
During the Fall of my junior year, I Co-Founded a software plugin called pe•ple with my friend and roommate, Han Wang. As a Co-Founder, our mornings were filled with customer emails, and as a startup, hearing from users was especially exciting and beneficial.
That being said, answering emails can be time consuming and energy draining. This is where an idea began to foster: automating customer emails with information from a central knowledge base.
To get initial feedback, the team (listed above) launched Dunder Email, a demo trained on information related to Dunder Mifflin from "The Office". They learned a great deal, but most importantly, how customers interact with a company's knowledge base. After months of hard work, Promptly was created!
Promptly helps support teams put customer service on autopilot by generating replies based on your knowledge base. We built Promptly with every intention to ease the jobs of customer support reps at companies, big and small.
Start automating in just 30 seconds:
📧 Login with Gmail
📄 Connect your knowledge base (website, documentation, etc.)
✨ Watch the magic in action – automated email replies with the click of a button
How can we simplify replying to emails & receive feedback from users about drafts?
I was brought on during a transition period when the team needed help on some design initiatives that were informed by client feedback.
The core functionality of Promptly allows users to draft customer support email replies in one-click. It is powered by GTP-3, an autoregressive language model that uses deep learning to produce human-like text. Our MVP only allowed users to draft a "Promptly Reply" using the context of the most recent email.
However, an additional use case emerged where users emailing within an email thread wanted to draft a "Promptly Reply" that took the context of the entire thread into consideration.
Additionally, we found another use case that revolved around training the knowledge base: users wanted ways to improve replies drafted by Promptly, but there was no feature that afforded users the ability to provide that feedback.
1. Users want the ability to draft responses that take information from an entire email thread into account
2. Users want to be able to provide feedback on drafted replies to train the knowledge base
Let's make some mock-ups!
The entry point for the "Draft from Thread" and "Feedback Report" features would be from the 'Reply' screen of an email. As a result, I mocked up the initial designs for both of these features on gmail screens.
Functionality to Read Previous Emails:
Ultimately, I decided to go with the exploration that includes the "Draft From Thread" CTA rather than a check box under the reply button. This is because:
- "Draft From Thread" reduces the amount of clicks needed to form a reply
- Having a CTA next to the reply button is familiar in terms of button layout
- "Scan Previous Emails" copy does not convey an accurate depiction of what the check box is actually doing
Selection Box Exploration:
- Check mark icon is more consistent with Google’s design system, versus a fill in
Generate Feedback Report for a Draft
- Smiley versus Thumbs Up Icon
- Smiley face is more associated with the feeling of “Satisfaction”, where as a thumbs up is more associated with “Approval”
- “How satisfied are you with this experience?” surveys often use smiley faces on a likert scale - these can be seen in public bathrooms/restaurants/venues etc.
- Aligns more with how you feel - can identify more with a face than thumbs up
results & metrics
Goodbye to monotonous customer service emails.
Talking to users, I was pleasantly surprised by how many times I heard, “I love this 'Draft from Thread' feature.” Promptly has done a great job of helping users craft helpful responses and save time for more important business tasks. I hope 'Draft from Thread' and 'Feedback Report' can further this by helping users optimize the way reply to customer service emails.
There are a lot more decisions and detail that went into this project than I mentioned here. If you'd like to learn more, please reach out!