Forethought

Discover

Overview

Forethought transforms the customer experience by infusing human-centered AI at each stage of the customer support journey.

Discover was a new Forethought product that I researched, designed, and helped launch from scratch within three months. I then used customer feedback to rapidly iterate on and enhance the first version of Discover.

My role

When I joined Forethought, the AI/ML engineers had already developed the majority of the NLP technology Forethought wanted to use to provide data insights for customers. My task, as the main designer on Discover, was to validate whether the insights were actually valuable to users, and if so, use them to design the first version of Discover. I worked with 1 PM and 8 engineers.

Problem & Goals

Customer experience admins and managers require lot of data cleaning and slicing to identify support process optimizations.

With Discover, our goal was to automatically categorize customer tickets into topics using NLP. Then, we can help them pinpoint inefficient areas of their products and support processes by providing intuitive data analytics.

Research & Data Validation

I conducted user research with 7 potential customers.

The ML team gave me a spreadsheet of data for each customer I talked to. By showing customers low-fidelity wireframes that used their real data, I was able to receive early feedback on the Discover concept.

Research Findings

  • Our selected insights interested users, but they would prefer the freedom to filter/sort all the data themselves.

  • All participants wanted to compare data over different time periods — this was the most valuable insight.

  • Our auto-generated topic labels made sense to users, but they noticed duplicate topics.

  • Users found the most value in the granularity of our topics.

 

After synthesizing findings, I shared them with the team, especially the ML team for feedback around topic generation and data analysis. Then, due to the short timeline of the project, I started exploring and refining high-fidelity designs.

Explorations

Given that Discover was a new product, I explored varying directions with the goal of providing users with a dashboard/summary view, as well as a view where they can filter/sort all the data themselves. I also explored different ways to allow users to re-visit data they want to monitor.

 

Final prototype demo

Bonus: Design System work

Throughout my process of designing Discover, I updated and/or created new design system components. Here are some highlights.

I worked with engineers to create a universal date picker, that was used across the Forethought platform (including Discover).

I worked with engineers to create a universal date picker, that was used across the Forethought platform (including Discover).

A universal filter button component was implemented during post-V1 iterations.

A universal filter button component was implemented during post-V1 iterations.

Before Discover, Forethought's side nav was always expanded, and could not be collapsed. Because Discover's dashboard required more horizontal screen space, I made the side nav collapsable across all of Forethought.

Before Discover, Forethought's side nav was always expanded, and could not be collapsed. Because Discover's dashboard required more horizontal screen space, I made the side nav collapsable across all of Forethought.

Results

Within the 3-month time frame, we launched and piloted Discover with several customers, which was our main indicator of success. Releasing the first version of Discover allowed us to collect customer feedback for post-V1 iterations (as shown above) and laid the groundwork for future versions of Discover.