case study · Mookti
I redesigned the first screen of Mookti, an AI study app, so new learners could start working with Ellen, its AI tutor, before making an account. After the new screen shipped in early September 2026, the median new account reached its first real action in about 1 minute, down from 3 to 6 minutes.
Mookti pairs students with Ellen, an AI tutor who helps them plan, learn and stay on track. I lead its design, from the logo to the screens.
Mookti's first onboarding was too long and too restrictive. The old welcome screen offered about 12 choices, and most guests left without clicking any of them.
That first screen is where people decide. Since launch in August, 65% of guest visits have lasted under a minute (1,930 of 2,956), and only 16% did anything at all.
My goal for the redesign: get every guest to send Ellen a first message, without making them wait, sign up or read first.
Three ideas shaped every decision that followed.
1
She is the product, so she belongs front and center on the first screen.
2
Everything works before you make an account. The account is how you keep what you made.
3
Real content and real interactions do the explaining, not instructions.
The design took four rounds before launch and one after, and two of the most important changes were cuts.
Both started with Ellen asking "What are you learning right now?" A, an Ellen takeover, made the chat the whole first screen. B kept the dashboard but gave Ellen one dominant card and demoted everything else. I also prototyped a connected flow that started in A and landed on a B-shaped home. I chose B because A felt too restrictive.
Ellen sat front and center with three suggestion chips. A "Keep what you grow" card invited people to make an account, and an "Or just explore" row offered sample content: a grove, a learning path, flashcards and a pretend week.
A "First steps" checklist (0 of 4) guided people at first. I wasn't sure it helped, so I tried a "How Mookti works" accordion explaining Groves, Ellen and Paths instead.
It repeated what the explore cards and Ellen already showed. I removed it, enlarged the "Keep what you grow" card and added a small flip-through preview of plans, free tier first. This version shipped on Sep 7. It opened with "Let's get you moving" and told people they could try everything before making an account.
After launch, I took the same idea into each page: no tour up front. Short guide text sits in the empty areas where people add content, starting with Groves.
New accounts reached their first real action faster.
Other signals from the first eight weeks:
The weekly groups are small, 44 to 138 people, so I treat this as an early signal rather than proof.
Onboarding is one part of Mookti's design. I built its whole visual language, from the first logo sketches in 2024 to v4.
Mobile is the next problem. 80% of guests arrive on a phone (2,380 of 2,956), but desktop guests did something at twice the mobile rate (28% vs 13%) and made accounts about 8 times as often (20% vs 2%).
Mookti v4 launches in mid-October with Goals replacing Groves, so the page-level onboarding pattern moves to Goals next.
What I learned: Cutting mattered as much as adding. The checklist and the accordion both explained what Ellen and the explore cards already showed, so removing them let learners experience Mookti for themselves. Ellen and the sample content guide, and learners discover the rest by doing. I plan to bring the same approach to onboarding for Goals in v4.