Problem Statement
Product Background
Project Overview
Discovery
Define
Ideation
Prototype
Testing&Refine
Final Design
Reflection
You.com is an AI assistant that provides access to diverse AI models in one place and delivers responses with citations.
You.com's goal is to enhance user productivity and access to information. As Richard Socher, founder of You.com, states: To give people answers more quickly, make them more productive, efficient, and well-informed, with better privacy.
Diversity of AI Models, Personalization
As AI continues to transform industries, designers must keep pace with emerging trends. To prepare for this shift, I attended AI conferences and meet-ups, gathering insights to inform new design opportunities. These experiences helped me build a foundation for tackling AI-driven design challenges.
At the Mindstone Toronto AI Meetup, held in a historic church, I met David Li, a senior software engineer at You.com. What intrigued me about You.com was its open-source platform with a strong foundation yet immense potential for growth. Their openness to feedback and experimentation made it the perfect sandbox for exploring as a designer.
Product Designer
You.com
Product UI Redesign
Revamped User Flow Prototype
Figma,
Figjam,
You.com AI
Through 10 tester recording samples, 7 users spent over 50 seconds on the dashboard clicking different labels and buttons. Some of them quickly ran out of the free generation opportunities and quit.
The test report revealed that the dashboard overwhelms users with options, leading to confusion and hesitation. Additionally, with limited free trial times, users may struggle with mismatched AI models, resulting in suboptimal answers and frustration. This reduces the likelihood of users purchasing the Pro subscription to explore more models.
During the testing task of finding the optimal model for "adpt this copy to our brand voice" and use it to generate the answer to this prompt, 5 users found the "Brand Voice" model under "By You.com" section but failed to navigate to a new chat by clicking the card, while 8 users didn't scroll down to the "Explore" section to find out the clickable prompt.
The model introduction page overly emphasizes crediting the team and redundantly introduces a single tool, cluttering the interface and neglecting to highlight key features effectively. This can result in a less streamlined, intuitive experience and may alienate users from exploring new models.
The unique value proposition (UVP) of You.com lies in its ability to provide users with access to diverse AI models so that they need only to subscribe to one platform to be cost-effective. However, growing user loyalty isn’t just about offering tools—it’s about helping users adopt the right tools for their specific tasks. If the product focuses on respecting human-intent, guiding users seamlessly leveraging the most relevant AI models, it can foster trust, loyalty, and repeat engagement.
Users don’t have the time to explore every tool in depth. They need a quick, intuitive overview—a sort of "elevator pitch"—to help them identify the right tool for their needs, making it easier to return for future tasks. This approach saves time and builds confidence in the platform’s value.
The needs of these two groups were vastly different, with power users seeking advanced features and customization, while new users required simplicity and clear guidance to navigate the platform effectively.
The inspiration for this approach came from AI platforms like Perplexity, Dot, and ChatGPT, which prioritize simplicity by grouping related features and automating tasks for easy access. The idea is to organize the workspace, making relevant tools available based on user intent, reducing cognitive load, and enhancing productivity.
The prototype introduces a smart agent selection feature that automatically predicts user intent and narrows down tool choices. By analyzing user input and context, the system delegates tasks to the most suitable AI model, eliminating the need for manual selection. This not only reduces decision fatigue but also highlights the advanced capabilities of the AI platform.
"Way-finding" serves as an analogy for improving navigation within the platform. By highlighting the capability of agents concisely and allowing customized organization for related questions, the platform creates a more intuitive structure. Like way-finding for travellers, these features simplify navigation, create landmarks, and enable users to focus on their goals, not tools.
Do it all the way, or don’t do it at all. If the goal is to create a tidy and streamlined UI, it’s essential to fully commit to removing unnecessary or confusing information from the user’s path.
During testing, some users found the remaining labels on the Threads cards confusing after I moved the labels above the prompt bar. This feedback highlighted the importance of consistency in design choices—partial changes can lead to confusion.
Surface-level model labels proved unnecessary—new users found them confusing, as they lacked the context or familiarity to make sense of them, while power users find them redundant since they won’t provide more information. Hiding the labels for the homepage caters better to both user groups.
To balance business needs and UI unification, I extract Trending Inquiries hiding in Agent page to an inspiration pool for all users to explore. This feature presents AI models as the "main character." Users are aware that they’re here to explore, so labelling becomes natural. Meanwhile, the product gets a chance to showcase its powerful multi-model service. This approach seamlessly integrates promotional elements with informative design, satisfying both business objectives and user needs.
User testing of the revamped You.com prototype showed promising results. Participants experienced 40% less confusion navigating the interface, thanks to the streamlined design and smart agent selection. The inspiration pool also boosted engagement, with 75% of users exploring multiple AI models. These findings suggest the redesign has a strong potential to enhance user satisfaction and efficiency.
I’d dive into designing and testing some of the branch features I couldn’t fully explore, like what happens when users search through their generated history or when they want to view the sources behind a generated answer. These are important moments that could build trust and make the experience feel even more seamless. I’d also love to take a closer look at the differences between the mobile app and desktop browser versions, making sure the design feels intuitive and optimized for each platform. These would be exciting areas to refine further and take the project to the next level.