Summarizing workflow & survey responses to save HR leaders time @ Assembly
Leverage AI's strength in summarization to help company leaders make the most sense of the data that goes into Assembly.
Objective
User, Product, and Business Goals
As a user and an HR professional…
I have set up a variety of culture and comms initiatives for my organization, and I need an easy way to synthesize the information that is flowing through the platform.
The product should…
Guide the user to valuable insights on their data. Getting to a summary should feel effortless, conversational, and magical.
Role and Responsibilities
Work with the Product Manager and Engineers to understand technical complexity. Conducted research, user testing, and UI/UX design for the feature.
Limitations
Perfecting function calling on our internal data would take too much time to launch, so we needed to guide our users through a series of questions before the summary to guide the AI on what to summarize without making the experience feel clunky or overwhelming.
Integrating with ChatGPT means paying for tokens on the amount of data that needs to get summarized. Another limitation was to find a creative way to cap tokens-per-query to keep our costs down.
The Feature
Dynamic Chat Experience
Chat-based experience that reveals questions dynamically based on the previous answers. This works perfectly to guide the user through set choices, and it allows us to track how much data is getting fed to the AI API. By asking a series of questions, we limit the amount of data based on a final period of time to summarize.
Portable Pop-out Window
This flexbile window has many use cases within Assembly, but it ultimately allows users to create and compare multiple summaries with different focuses, and continue to browse the rest of the application while maintaining context on their insights.
Outcomes
Since the release of this feature, the Sales and Account management team have been able to upsell accounts based on the usefulness of this feature. Upsold deals who use this feature bring in 25% more revenue, and average per-seat price has been going up a 10 cents MoM for the past 4 months since release.
Some popular use cases we've seen:
Creating meeting agendas
Summarizing 360 Review feedback and providing next steps
Summarizing a month's worth of Daily Standup posts to help new team members get up to speed
Exploration
This feature took a lot of iteration to figure out how to naturally dovetail structured questions into an AI-based response. Here are some approaches I tried before landing on the full chat experience:
