Korbit , a personal AI Tutor
Designing an engaging experience for an AI tutor using reward systems, progress dashboards, and re-engagement emails.
About the company
Korbit AI is an EdTech platform that helps students and early career professionals start or transition to a career in Data Science.
It also supports large enterprises by training their employees and potential hires in Data Science and related fields. The website features an AI-powered conversational chat interface that acts as a 'personal AI tutor,' showing videos, asking questions, and responding to students' answers. Video about Korbit, the AI tutor
Note: This project was completed before the advent of ChatGPT.
The Korbit AI interface consists of three parts -
Project Details
Team
Kritika Sony - Interaction Designer
Tommy Delarosbil - Senior Product Designer
Stephane Robert - Director of Software Engineering
Duration
6 months
My Role
User Research
Ideation
UI Design
Design handovers
Client
Korbit AI
A Canadian EdTech company that created an AI tutor (before ChatGPT got big!)
01
The Website
This page consists of relevant information about what the AI tutor does. The user signs up and is led to the onboarding page.
02
Onboarding
The onboarding flow evaluates users' current skills and their learning preferences.
For example: "Do you like learning by coding or by doing exercises?" and "Are you interested in particular topics?"
03
AI Chat Interface
This is where the user starts to interact with the AI Tutor.
It provides context for the topics, gives you relevant videos to watch, asks questions and also answers users' doubts.
My work mainly revolved around the website and the AI Chat Interface. There were other designers in the team who owned the onboarding experience.
Problem
Korbit AI had really great learning outcomes. However, the company suffered from low engagement rates. The data scientist of the company presented the following statistics to the team:
Nearly a third of users did not participate at all and many others had minimal activity.
A significant portion of users have very low active time, with many users clustered around the lower end of the active time spectrum. The 25th percentile of active time is zero, and the median (50th percentile) is only 31 units.
How might we enhance the learning experience for new data science students on Korbit to boost engagement and retention?
Solution
01
A reward system + encouragements within the AI chat interface
02
A progress and rewards collected dashboard
03
Reminder emails for when the user forgets to log back in
01
Reward System within the AI chat interface
We created an extensive rewards system that was injected within the AI chat interface and was based on the achievements such as reaching a certain lines of code or watching certain number of minutes of lectures, making progress on problems, practicing regularly for a set number of days, trying again and again till they get it right, etc.
02
Progress Dashboard in the Profile page
These rewards are displayed on the profile as a dashboard where users can track their achievements, see the additional rewards they can earn, and experience a sense of accomplishment. This system reinforces the feeling of leveling up as they learn with Korbit.
Version 1: Light Mode
Version 2: Dark Mode
03
Service Design: Reminder Emails to bring user back to the platform
Additionally, I designed and implemented emails that encourage users to come back to the platform, if they have not visited the platform for a while.
We Miss You at Korbit. 🥺 Let's Get Learning Together!
Hello Kritika,
It's Korbit here, your friendly AI tutor! It looks like you started learning Machine Learning but haven’t jumped back in yet. Let’s continue your journey together! There's so much we can learn together!
Here’s what you’re missing out on:
Personalized Lessons: I’ll tailor your learning experience just for you!
Interactive Fun: Engaging activities and challenges to make learning exciting.
24/7 Support: I'm always here to help you, no matter when you need it.
Let's pick up where we left off and continue your learning journey. Click below to jump back in:
Login
If you have any questions or need a bit of help getting started, feel free to reach out to our support team at support@korbit.ai.
Looking forward to seeing you soon!
Your friendly AI tutor,
Korbit
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Get the most out of your Korbit Experience
Hello Kritika,
We hope you’re enjoying discovering our learning platform! If you need any assistance along the way, please don’t hesitate to reach out at support@korbit.ai.
To get the most out of your learning experience, we recommend using Chrome or Firefox on your desktop or laptop. For the moment, we don't guarantee compatibility with other browsers or mobile so some issues might occur when using unsupported configurations.
Looking forward to hearing about your experience!
Login
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
Your friendly AI tutor,
Korbit
Resume Learning
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
🌟 Ready for the Next Step in Machine Learning?
Hi Kritika,
It’s Korbit again, your friendly AI tutor! I hope you're excited as I am to continue our Machine Learning journey. You’ve already made a great start—let’s keep the momentum going!
Next, we’ll explore:
Neural Networks: Dive into the building blocks of AI.
Data Preprocessing: Learn how to prepare data for machine learning models.
Hands-On Practice: Engage in activities to apply what you’ve learned.
Ready to continue? Click below to jump back in:
Hi Kritika,
It’s Korbit again, your friendly AI tutor! I hope you're excited as I am to continue our Machine Learning journey. You’ve already made a great start—let’s keep the momentum going!
Next, we’ll explore:
Neural Networks: Dive into the building blocks of AI.
Data Preprocessing: Learn how to prepare data for machine learning models.
Hands-On Practice: Engage in activities to apply what you’ve learned.
Ready to continue? Click below to jump back in:
Resume Learning
If you need any assistance or have questions, just let me know. I’m always here to help!
Looking forward to seeing you soon,
Remember, I’m here to support you every step of the way. If you have any questions or need help, feel free to reach out!
Happy learning,
Your friendly AI tutor,
Korbit
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Process
Understanding goals
Business goals
Create a north star vision of the company for 2 to 3 years after the start of this project.
One of the company's KPIs for that year was to increase the engagement rate to an average of 30 minutes per user session or complete 2 learning units (approximately 30 mins) per session.
Design goals
Understand the reasons behind why the users are dropping off, and to find and fix them
Research with user data
I worked with the data scientist to learn more insights with the data, I divided the data and defined active users, power users and not engaged users.
Learning about the highly engaged users made us question — What drives some users to stay for over 30 hours and complete as many as 50 learning modules, while others leave the platform almost immediately?
To better understand this, we decided to conduct Semi-structured User Interviews and Website analysis.
Highly Engaged Users
The user base had a small group of highly engaged users which provided an opportunity to understand what drives their behavior.
User Active Time Analysis
The data showed that most users had low active times, while a smaller number had significantly higher active times.
UX Research
Semi Structured Interviews
Considering the user data analytics, we decided to recruit users who had completed at least 3 learning units on Korbit, as they had sufficient experience to provide insights on what worked and what didn't.
Also, the users who spent several hours on the platform could help us understand what kept them engaged.
I recruited 5 Korbit learners by sending out mass emails filtered to meet the required criteria.
Alongside the Product Analyst and Senior Product Designer, I took turns moderating the session and taking notes.
The interviewees were offered a $30 gift card for participating in a 30 to 60-minute interview.
Website analysis
The users who left almost immediately would not have enough experience with the Korbit platform to provide useful feedback. For them, we decided to analyze the initial friction points before they even reached the chat interface — the Korbit website and onboarding process.
For website performance, I analysed Mouseflow (a website analytics tool that tracks and analyzes user interactions) using heatmaps and suggested data driven iterations.
After discussing these iterations with the leadership, I implemented several of these of Webflow, the platform on which our website was built.
The onboarding process was part of a separate project that I will not discuss here.
Main takeaways:
From the interviews, we learned that users who interact with AI chat interface for hours and continue learning have learned "how to talk to the AI". Initially, they found the AI frustrating too but once they trial and error'd their way through the initial set of challenges, they found the Korbit platform super useful.
The major driver of engagement were the quality of the AI responses. The users had varied opinions on how useful/not useful they were.
This led us to deep dive into the specifics of the AI chatbot responses. We decided to analyse specific instances and this time using observational studies using Hotjar (a website analytics and feedback tool that provides insights into user behavior to optimize user experience) session recordings
Observational Studies - recorded user sessions
I watched 43 recorded sessions to note which exercises were helpful, not helpful and to guage reasons for learners dropping off.
While this was created primarily to understand user behavior, this analysis tremedously helped the engineering team understand where the AI is answering incorrectly and prioritize problems to work on.
Peeking into how are others solving this problem
Now that we had a good understanding of our platform and the specific problems it was facing, we decided to look at other platforms which have succeeded in driving regular engagements to get some inspiration. We looked at Duolingo, an app known for its experiments with engagements. Additionally, we also looked at games that encourage users to come back to the platform on a regular basis such as Subway Surfers and become a 'habit' for users.
I benchmarked Duolingo and the game Subway Surfers for habit. I marked their rewards, encouragements, challenges, both within the sessions and after completions.
Synthesis
User Personas
To align the leadership and the company vision in general and to represent the key user groups, we designed personas based on the insights from user research. These personas helped the team keep the users in mind throughout the design process.
The first persona is Andrew, who just wants to get the work done and is extrinsically motivated with certifications.
The second persona is Charlie, who prioritizes his work and learning and is using Korbit to upskill for both personal and professional reasons.
User Journey Mapping
It is at this stage that we decided that learning with Korbit is not just about learning but also about the experience of learning. For this we mapped our user journey map for a happy path for a first time user
This journey map made us think about the problems in two ways for potential areas of improvements:
01
I proposed and presented the smallest, most feasible solutions that could be immediately implemented to significantly impact the user experience.
02
We discussed what can we do to enhance the experience of learning with Korbit in the long run (2-3 years later) from the start of this project
For 01, we quickly prioritized and implemented several small but impactful solutions. These included improved and friendly copywriting, providing more explanations and checks, informing users at each step about the AI's responses and what to expect from the chatbot, and adding a progress bar.
For 02, we conducted ideation workshops involving employees from various teams, including engineering, design, product, and content management. Since different teams prioritize problems differently, it was crucial to ensure everyone's voices were heard to develop effective solutions. (See Ideation Phase)
Ideation
Future Workshop and Metaphorical Design
We began by conducting a Future Workshop with employees from various teams, including engineering, design, product, and content management. We explained the problem statement and initiated the first round, where everyone had to propose one idea to improve and make the Korbit platform more engaging—no restrictions, wild ideas encouraged. In the subsequent rounds, each team member built upon each other's ideas, continuing this for five rounds.
At the end, we voted on the top ideas.
These were the top ideas from the team:
The Korbit Solar System
Imagine each learning module as a planet in a solar system. As you learn new skills, you add more planets, expanding and enriching your educational journey. Each learning module becomes a distinct planet in this growing system.
Korbit as a chef in a restaurant
Imagine you're the chef at a restaurant. As you learn new skills, you add more dishes to your specialties, expanding your menu and showcasing your growing expertise.
Korbit as a human
Imagine Korbit teaching you just like a human. This AI interacts with you as if it were a real person, answering your questions directly. It's like having a personal expert available 24/7, guiding you through your learning journey.
Korbit as a forest ecosystem
Imagine your skills as plants in a forest. As you learn new skills, you add more plants, making your forest greener and richer. Each new skill you acquire contributes to the growth and vibrancy of this flourishing forest.
Explorations
These were some very interesting ideas. These were all ideas in text so to get a better understanding of the team's thoughts, we decided to visualize these ideas:
Exploration 1: Korbit Solar System Mockup
We thought about what would an exploration truly look like? We created a bunch of scenarios which visualized the idea. The firtst exploration was a Korbit Solar System in which the user is a planet and there are more planets and skills that the user can add to their solar system. The good times and encouragements are marked by Aurora Borealis and the bad times are shown by hurdles such as meteor showers. We also added a social element where users see other learners as planets.
Exploration 2: The Plant Ecosystem
To focus on the 'growth' side of progress, we thought what could be better than having a plant grow. From a seedling to a tree and eventually an entire forest.
Meeting with Stakeholders
We met with the leadership and conducted a meeting to guage their thoughts on both these ideas.
We learned that building something like these are unattainable at the moment given the time, effort and limited engineering capacity. However, this set a good direction on where we would like to see Korbit 2-3 years from the start of this project.
Asking 'why' these ideas were appealing
To get to the core of why these ideas are so appealing, we went back to our research, notes from meetings and learned that the reason that these ideas were so appealing because
they showed progress — such as adding planets and helping a plant grow
they showed a sense of achievement — creating a solar system and creating a forest
they were engaging — these provided an interactive experience with something they could visualize
From these, we realised, we can still create a platform that fulfills all of these requirements by creating:
01
A reward system + encouragements within the AI chat interface
02
A progress and rewards collected dashboard
03
Reminder emails for when the user forgets to log back in
Final Designs
01
Reward System within the AI chat interface
We created an extensive rewards system that was injected within the AI chat interface and was based on the achievements such as reaching a certain lines of code or watching certain number of minutes of lectures, making progress on problems, practicing regularly for a set number of days, trying again and again till they get it right, etc.
02
Progress Dashboard in the Profile page
These rewards are displayed on the profile as a dashboard where users can track their achievements, see the additional rewards they can earn, and experience a sense of accomplishment. This system reinforces the feeling of leveling up as they learn with Korbit.
Version 1: Light Mode
Version 2: Dark Mode
03
Service Design: Reminder Emails to bring user back to the platform
Additionally, I designed and implemented emails that encourage users to come back to the platform, if they have not visited the platform for a while.
We Miss You at Korbit. 🥺 Let's Get Learning Together!
Hello Kritika,
It's Korbit here, your friendly AI tutor! It looks like you started learning Machine Learning but haven’t jumped back in yet. Let’s continue your journey together! There's so much we can learn together!
Here’s what you’re missing out on:
Personalized Lessons: I’ll tailor your learning experience just for you!
Interactive Fun: Engaging activities and challenges to make learning exciting.
24/7 Support: I'm always here to help you, no matter when you need it.
Let's pick up where we left off and continue your learning journey. Click below to jump back in:
Login
If you have any questions or need a bit of help getting started, feel free to reach out to our support team at support@korbit.ai.
Looking forward to seeing you soon!
Your friendly AI tutor,
Korbit
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Get the most out of your Korbit Experience
Hello Kritika,
We hope you’re enjoying discovering our learning platform! If you need any assistance along the way, please don’t hesitate to reach out at support@korbit.ai.
To get the most out of your learning experience, we recommend using Chrome or Firefox on your desktop or laptop. For the moment, we don't guarantee compatibility with other browsers or mobile so some issues might occur when using unsupported configurations.
Looking forward to hearing about your experience!
Login
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
Your friendly AI tutor,
Korbit
Resume Learning
🚀 Let’s Dive Deeper into Machine Learning with Korbit!
🌟 Ready for the Next Step in Machine Learning?
Hi Kritika,
It’s Korbit again, your friendly AI tutor! I hope you're excited as I am to continue our Machine Learning journey. You’ve already made a great start—let’s keep the momentum going!
Next, we’ll explore:
Neural Networks: Dive into the building blocks of AI.
Data Preprocessing: Learn how to prepare data for machine learning models.
Hands-On Practice: Engage in activities to apply what you’ve learned.
Ready to continue? Click below to jump back in:
Hi Kritika,
It’s Korbit again, your friendly AI tutor! I hope you're excited as I am to continue our Machine Learning journey. You’ve already made a great start—let’s keep the momentum going!
Next, we’ll explore:
Neural Networks: Dive into the building blocks of AI.
Data Preprocessing: Learn how to prepare data for machine learning models.
Hands-On Practice: Engage in activities to apply what you’ve learned.
Ready to continue? Click below to jump back in:
Resume Learning
If you need any assistance or have questions, just let me know. I’m always here to help!
Looking forward to seeing you soon,
Remember, I’m here to support you every step of the way. If you have any questions or need help, feel free to reach out!
Happy learning,
Your friendly AI tutor,
Korbit
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Our Blog • Contact Us • Unsubscribe • Terms and Conditions
Impact
Re-engagement Emails: "I designed and implemented re-engagement emails, resulting in a 300% increase in user retention within the first month."
Future Workshop and Metaphorical Exercises: "The future workshop and metaphorical exercises defined the company's north star, steering conversations around its vision and long-term goals."
User Feedback on Learning Experience: "Users reported a more intuitive learning experience, with an 85% satisfaction rate. Feedback highlighted enhanced interactions with the AI tutor, making learning more engaging and personalized."
