Case Study

Bridging the gap between UX and AI at enaible.

Abstract

Enaible is an artificial intelligence platform that leverages pre-existing systems data (O365, salesforce, etc.) to measure and improve productivity in the workplace. I led the research, design, and development of a new product experience that resulted in a 64% improvement1 in overall customer product satisfaction, helping to solidify over $1.5 million in proposed contract renewals.

a

Overview

Background

In 2020, enaible completed research and development on its core AI engine and wanted to create a new product experience to deliver its insights to end users. While a prior product experience existed, it diverged from the latest AI engine and didn't support current business goals. In early 2021, enaible hired me to lead the research and design of the new product experience as well as the entire frontend development process.

Problem

We discovered early on that the outputs from the core AI engine were difficult for users to understand. Furthermore, early client feedback seemed to suggest a negative sentiment towards AI-based management. It was clear we needed to determine how to bridge the gap between artificial intelligence and user centered design.

Interesting Fact:

Machine learning algorithms are often considered to be “blackbox” as the input and output values are known, but the steps in between, or the means in which it arrives at its output values, are unknown. This has lead to negative sentiment towards certain forms of AI, creating additional fields of research like XAI.

Learn more here
Objective

After various conversations, we identified the following objective: research, strategize, and design a progressive web app (PWA) experience that helps employees improve workplace productivity through understandable AI insights while increasing positive sentiment towards AI-based management.

Research

The Kickoff

As we began the project, we conducted the first portion of an in-depth qualitative study that collected data from over 25 users in focus-group discussions and individual interviews. We observed individuals as they used the existing interface and followed up with a guided survey experience that helped us to better understand how our users felt about our AI and the existing product experience.

Partial example of the guided survey

The results of this study helped confirm our initial problem hypothesis, namely, that the AI lacked sufficient credibility. Furthermore, we discovered that there was a large percentage of participants who were concerned that the platform was creating a form of professional "big brother" and that their data could be used against them at the company's discretion. We had many other key takeaways such as:

Key Findings:
  • 82% had difficulty connecting our productivity insights with actual work habits.
  • 78% held negative sentiment towards AI-based management (AIBM).
  • 73% would have more openness towards AIBM provided their data was anonymous.
  • 62% reported wanting to know which KPIs and work behaviors impacted their productivity.
Identifying Users

After finalizing our initial study round, we moved on to identifying our primary user personas. Here are some examples of the personas we worked with below:

foo
Examples of some personas we worked with

Once we completed user research, we consolidated our findings into three guiding principals that felt best defined our new product vision:

Guiding Principals
  • Improve workplace productivity through explainable AI
  • Promote career growth & employee independence
  • Prioritize user anonymity & data privacy

Strategize

Sifting through the data

Now that we had data to work with, it was time to begin planning our redesign strategy. Since there were over 1,200 combinations of AI driven insights at our disposal, we began by utilizing an object-oriented user experience2 approach to prioritize and consolidate them. These insights were then matched with existing user pain points to help us create conceptual features and user journeys.

Time to prioritize

Once we had user journeys and clearly defined feature stories, we prioritized them using a traditional eisenhower matrix.

We ended up selecting 35 insights to deliver to users.

Phew! At this point we had a plan in place and were ready to move on to design.

Design

The results of our research and strategy phases yielded better results than we expected. It was time to put it to work in creating our designs. Before we began, we agreed on the following design guidelines to help us on our journey:

Visual Language
  • Vibrant and engaging
  • Minimal yet informative
Messaging
  • Motivational yet honest
  • Always focused on privacy
Design Architecture
  • Scalable/modular design
  • Prioritize accessibility
Starting with concepts

Now that we had our design guidelines, we worked on creating various wireframes to help us understand the information hierarchy and interactions of each feature. Many of the wireframes never made it to fully designed prototypes, but it helped save time in the long run.

Examples of the wireframes.
Hundreds of views

We spent a few weeks creating hi-fidelity prototypes, but needed to be sure that our developers were aware of every edge case. This meant creating hi-fidelity prototypes for each possible path in the user's journey.

Examples of the product flows.
Life through animation

We added animations on the most critical features of the platform to give life to the experience. You can see the fruit of our labor below:

Examples of the product animations.
Mobile prototypes and alterations

We also created mobile views and various iterations of the desktop design.

Examples of the mobile views.
Crafting our voice

A few recent studies2 suggested that first person narrative could help to improve user sentiment towards AI. We experimented with various points of view, but ultimately found the first-person perspective to be the most effective at crafting a narrative. We incorporated it into our onboarding to help frame the product's AI and our commitment to privacy.

Example of the onboarding experience.
The final product

Once we had our final prototype, we revisited our focus groups and conducted the second round of our in-depth qualitative study. The results were promising and indicated we were on the right track with a 64% improvement1 in overall product satisfaction! You can see an example of the finished prototype below:

Examples of the finished prototype.

You can view the finished prototype below.