Background
edX offers over 5000 courses and programs which can be overwhelming to navigate. Learners do not have guidance for finding the best learning options for them apart from information directly on course pages. This puts pressure on our support staff or worse, can lead customers to either being unsatisfied with their choices or simply going elsewhere for their learning.
With the development of GPT-4, we identified an opportunity to wield artificial intelligence to provide a chat tool that could guide users to the learning content they are looking for. Xpert Chat is an AI-powered intelligent chat service that helps learners find the best learning options for them and navigate the plethora of learning content that we offer.
Approach
Although AI chat tools are a new development, there are a plethora of chatbots and digital assistants that are familiar to many of us. This was where I started, by collecting examples of previous chatbots and conducting an analysis of their successes and pitfalls. I conducted interviews and concept testing to gauge user expectations and experiences with chatbots. Using these learnings, I designed Xpert, an AI assistant for edX learners.
Business goal
We believe that a chat tool that guides to the best fit learning content will increase course enrollment, account registration, and retention because users will be able to more easily find what they need, be satisfied that they have found the best option, and have trust in edX as an education provider. Xpert will also provide a quicker start by helping learners jump right into their courses or programs.
Metrics
1. Increase edX account registration
2. Increase chat click through rate
Discovery
100's
of data points analyzed
Given that this was edX’s first instance of a chat tool I wanted to be thorough in our research efforts to ensure that we were creating the best possible experience for our learners. Specifically, I wanted to make sure to avoid the pitfalls of previous chat bots, such as having repetitive, script based responses and generally being unhelpful.
The first steps of my research was to do an exploration of past & existing chat bots and a literature review into chatbot and conversational AI best practices. I ran AB testing experiments on the edX website to learn what types of features users prefer. Then, I created a customer journey map to better understand user behavior and pain points during their course discovery. I conducted user interviews and prototype testing to learn about user behavior and validate my concept designs.
Historical chatbot references
Competitor references
Analysis
Affinity mapping
Research insights
Solution Design
Proactive Message
I designed a proactive message that pops up over the Xpert chat button when users first come on the homepage to increase open rates. This message led to a ~48% increase in chat open rates.
Animated chat launch button
I illustrated and animated a "graduate" character to use as a chat launch button, leading to a 30% increase in chat open rates.
Outcomes
The launch of this work then led to the development of two other instances of Xpert AI, which I led the design of as well: Xpert Learning Assistant for edX Courses and Bootcamps.
