Exploration 1: Faces on soft shapes
I started with faces drawn on soft, irregular shapes. Each had eyes, a nose, and a mouth drawn in loose lines. The founders felt they read too human and asked for something more AI forward.
Tenmin
Reshaping onboarding around an AI tutor learners never met
Tenmin is an AI-powered language learning app built around speaking rather than memorization. Learners get real time pronunciation feedback, personalized lessons, and conversation practice that simulates real interactions.
I led design and prototyping for the onboarding flow, working closely with the founders, the PM, and two other designers. The redesign introduced Tenmo, the AI tutor, before the app asked learners to sign up. I also redesigned the review home screen so learners could see what was waiting before starting a session.
-
Role
- Product Designer
-
Timeline
- April — July 2025
-
Collaborators
- Nadia Lee, Mauricio Rivero Pooley, Juna Han
-
Focus
- Interaction Design, Visual Direction, Prototyping
Problem
Meeting the tutor too late.
Tenmin’s AI tutor, Tenmo, is the reason the product exists. Learners did not understand who it was until their first lesson began.
The original onboarding asked learners to sign up before it explained anything. Five screens of taps and typed input followed. The AI tutor only appeared in the lesson. Learners reached their first conversation without knowing who they were about to talk to. The first thing that conversation asked of them was to speak out loud.
Discovery
What learners did not know.
I looked at onboarding across five language learning apps. Some rushed learners to sign up as fast as possible. Others opened with a list of features.
I also interviewed two learners. Both said they would have felt more at ease in the first lesson if they had met Tenmo beforehand. That changed how I framed the problem. Learners did not need a better explanation of what Tenmo could do. They needed to meet it first.
Character Explorations
Designing a tutor that learners could meet.
Before the redesign, the AI tutor had no face and no name. Tenmin’s brand had to be polished enough to be credible as a learning tool. It also had to be warm enough that talking to it did not feel transactional. The founders wanted the AI tutor to carry both, looking like AI while still reading as human.
The team shaped the brand direction together. My part was carrying it into onboarding, where Tenmo made its first impression.
Exploration 2: An orb with eyes
I moved to an orb with eyes, designed in Figma. The orb kept the eyes, which gave it a sense of attention, and dropped the rest of the face.
User Journey
Reordering the path to the first lesson.
The proposed flow opened with an introduction to Tenmin, social proof from other learners, and what learners would gain. Returning learners could log in from the first screen and skip the rest. New learners then chose a language and a learning goal, and set their name by typing or recording it. Tenmo appeared next and introduced itself. Sign up came last, just before course preference and the first lesson.
Solution
The redesigned flow.
I reordered the onboarding flow and added two new steps, an introduction to Tenmin and Tenmo’s reveal. I also added a voice option to name setup and produced the illustrations used across the flow.
Recording your name
The original flow collected names through a keyboard. I kept typing and added the option to say it instead. The recording was meant to let Tenmo greet learners by name at the start of each lesson, using their own pronunciation. It also gave them a low stakes chance to use their voice before the first lesson asked them to speak.
Meeting Tenmo
Tenmo appeared on its own screen after name setup. The orb winked as a voice said “Hello, I am Tenmo.” The greeting was the same whether learners typed or recorded their name. Keeping one version reduced complexity.
Review home screen
I swapped the order of Phrasebook and Smart Quiz after a discussion with the Tenmin founders. They wanted to steer learners toward practicing their weak areas, so Smart Quiz came first. I also renamed Smart Quiz to Flashcards to match what it does. Flashcards showed a card count and time estimate. Learners could pick a session based on how long it would take.
Directing illustrations with AI
I produced the illustration set with ChatGPT. The founders and the team agreed on a soft 3D style with rounded forms and gradient color. Every asset followed that direction. After generating the assets, I refined them by adjusting their color by hand in Figma.
Trade-off
Putting value before sign up.
I placed sign up after Tenmo’s reveal so learners could experience Tenmin before deciding whether it was worth an account. Placing sign up earlier could have brought more learners to it before they left. My choice had a cost. Learners who had met Tenmo and recorded their name could hit a sign up screen and feel tricked.
However, a learner who has met Tenmo has a reason to sign up. A learner facing a sign up screen cold has only a promise.
Outcomes
What testing showed and what Tenmin adopted.
The team tested the redesigned flow with three language learners. I ran one of the sessions. All three could say who Tenmo was and what it would help them do before the first lesson began.
Tenmin shipped the illustration direction and the onboarding microcopy. The name recording did not ship. As far as I know, Tenmin prioritized other features ahead of it.
Reflection
What I would do differently.
I would make the log in option more prominent. It appeared as a text link on the first screen, so returning learners could miss it and start onboarding again.
