Easy Onboarding of AI Tool

-How Dr.lambda improved onboarding experience to keep new users
Role

UX research, UX design, UI design

Time

10 weeks

Team

2 designers, 1 researcher, 1 product owner, 2 developers

Stats

Team project

Shipped
Background
Overview

Dr. Lambda is an AI powered slides generation tool aims to help academic users to create slides more efficiently. However, with growing numbers of users, we spot users also giving up the tool after trail because of their worry of potential high learning curves of the new tool.

The aim of the project is to solve the problem of users' doubt for potential high learning curve for AI tool through an guided learning tour.

Design Process
Problems
Problem Statement

How might we provide helpful guidance from start to finish to assist users with academic background to create slides effortlessly and intuitively?

Solution Preview

Feature 1: More customized onboarding questions

It helps create more detailed user profiles and tailor the content generation to meet users' needs.

Feature 2: Step-by-step tour

It enhances user engagement with a seamless onboarding experience and leverages AI to generate slides.

Feature 3: Feedback collection

Collecting feedback helps the AI better understand user preferences and deliver more tailored content.

Research
Research Process

Listens to users

Talking to users and seeing their struggles help us to identify the questions and confusions they have.

Persona & Journey Map

New users feels more and more confused and finally disappointed

New users are unfamiliar with the AI slide generation workflow, which has caused confusion.

Research Insights

Understanding user pain points

Identifying key challenges users face to design solutions that address their needs effectively.

Answers don’t match purposes

Users want AI to retain their preferences instead of re-entering them every time.

AI workflow feels confusing

Users don’t understand the AI process, making it hard to complete tasks smoothly.

No way to share preferences

Users lack a way to provide input, preventing improvements to generated results.

Features

Registration survey for general purpose

Understand users' purpose upfront to avoid repeated input

Guided workflow simplifies process

Step-by-step guidance helps users understand different AI workflows.

Feedback loop refines results

Collect preferences to improve and personalize future results.

Design
Feature 1: Onboarding Questions

Collecting user data helps AI generates more ideal content

Users felt the content lacked personalization. To provide more tailored results, we now collect users' goals at the start of their journey to help the AI better understand their needs.

Feature 2: Guided Tour

Go through important steps with support

We designed a guided tour to provide users with a smoother first-time generation experience.

Feature 2 Testing

This is kind of too long for users

Through testing, I found that users gradually lose interest in the guided tour and want to quit early if their doubts have already been cleared.

Testing Insight

We should adjust the rhythm of onboarding and provide early exit points.

Feature 2 Improvement

Adding section breakpoints to have a better rhythms

We divided the tour into three sections to celebrate users' progress and give them a break. This turned the long journey into three shorter ones, with the option to skip.

Feature 3: Feedback Collection

Tell us how you think

I designed a questionnaire at the end of the onboarding journey to gather user feedback and improve AI performance.

Feature 3 Testing

We need to shorten the feedback form

Users find the feedback form too long, so we shortened it after consulting the developer on key data needs.

Testing Insight

We need to keep only essential questions and make them easy to answer.

Feature 3 Improvement

Reduce questions so users can quickly finish it

We kept only two quick-choice questions to ensure users feel no pressure.

Visuals
Design System

A scalable design system for future development

We developed a comprehensive design system inspired by Google Material Design and Ant Design, aligned with the company's branding, and revamped the UI to ensure consistency.

Takeaways
Learnings
1
1

Don't fall in love with your design

The first usability test was terrible. At the beginning, I felt sad for having such a bad experience for users, but later I find other senior designers and engineers have no hard feelings about the feedback. I learned to accept comments objectively, even for your own work.

2
2

To avoid learned helplessness

If users fail again and again, they would get helpless and stop trying anymore. Therefore, it's very important to help them get anticipated results to ensure they get wanted result.

3
3

Proper feedback within 0.1s

Proper feedback should happen within 0.1s after users' actions to tell users the system has received their request to let them be reassuring. If the request takes some time, they should be aware about the progress.