A Copilot to find a right AI tool
CosmoS


Overview
Problem
In today's AI-driven world, many individuals find it challenging to discover and effectively utilize an AI tool that meets their unique work and study requirements. Our target audience consists of those who like to leverage AI for their professional and personal lives but who feel discouraged by the amount of information and the time and effort needed for research, testing, and learning. How can we streamline the process of locating and using the right AI tools, making them more accessible to all?
Solution
CosmoS is a platform allowing users to get the most recent AI industry information and updates, and streamline the process of locating and using the right AI tools. Featuring a copilot and the experience of browsing while chatting, users can quickly navigate to the tools that specifically meet their needs and the related contents such as tool reviews, tutorials, relevant projects, etc. The platform solves the problems of overwhelming and dynamic info on AI tools, unpredictable/high learning costs and scenarios, and accessibility of AI tools.
My ROLE
UX Research
Prototype and Iterate 2
Presentation
Teammate
Chenghong Tang
(UX Researcher, Prototype Drafter)
Timeline
Iteration 1
2023.10-2023.12
Iteration 2
2024.6
Product
Copilot: Ask then Answer
To address the issue of users struggling to find the right tools, a copilot can provide tailored recommendations based on their specific needs. In addition to suggesting tools, CosmoS can also recommend the most relevant online tutorials to help users complete their tasks.

Tab Experience
Drawing inspiration from browser tab experiences, I integrate tabs into CosmoS to maintain organized functions while also creating an innovative and immersive environment for exploring and learning about AI tools and related information.

Connections between Copilot and Tool Pages
The tabs link the copilot with tool details and additional pages. If users are interested in a recommended tool on the chat page, they can click on it to open a new tab displaying the tool's detailed page, which features personalized content and the latest discussions. Conversely, if users encounter any confusion on the tool's detailed page, they can add that content to the chat page and pose their questions. This method helps users verify information through the copilot while fostering an engaging environment for exploration and learning.

Splitscreen If Preferred
Clicking back and forth between tabs could be frustrating in some cases. Users can adapt to their preferences by utilizing the split-screen feature. This allows them to concentrate on chatting or seamlessly switch between pages to enhance their overall experience.

Research
Persona

Name: Jane Wong
Age: 25
Family Status: Single
Economic Status: Poor
Hometown: Brooklyn, New York
Occupation: Musical Student
Working style: Multi-tasking
AI Usage: Primarily ChatGPT
Problem
Jane is approaching her graduation from music school and faces a busy schedule ahead. She needs to develop and present a prototype for her graduation musical within two weeks, but she urgently requires a draft script despite having a concept mind.
While is comfortable using AI, her initial experience with ChatGPT left her unsatisfied. She is determined to find a better tool to meet her needs, but as she navigates an overwhelming amount of information online, she feels discouraged and stuck in her work due to the time and effort required to find and learn new tools.
Interview: Do you use AI? What discourages you from using them?
We interviewed six individuals.
Here are their answers summarized:
Often High Learning Cost
""
Generations could be slow and suck
Analysis
Not all AIs are intuitive
Need learning with cost of time and effort
Mix of mature and immature AI tools in the market
I'd like to use AI but I don't know about them
Good AI lack exposure
""
Current AI Market Analysis
Breakdown of generation quality and learning cost

Insights from Interview & AI Market Analysis
We interviewed six individuals.
Here are their answers summarized:
Often High Learning Cost
""
Insights: What we can do
Make the learning process less painful
Generations could be slow and suck
Be selective and expose the good ones
I don't use AI that often
Get the right tool ready for specific needs
""
Prototype Iteration 1
Community Building?
Our initial idea is to create a shared community with learning materials, self-explorations, forums, and AI recommendation list for users to get the right tool and make the learning experience less painful.
We hope the sense of community could help users more engaged and motivated with the researching and learning.
Below is our first prototype.

less painful learning process
Shared tutorials and projects

Find the right tool
Exploration Mode


Selection and Exposure
Exposure and Engagement
Tool community subscription
Review Session
Usability Test


Insights
-
The current product focus is distracting.
-
Community building could be an off track idea. We should focus on the tool recommendations and shorten the learning cost. People need a straightforward AI tool recommendation.
Prototype Iteration 2
Copilot Experience Reimagined
Product Structure Experiments & User Testing
Users are seeking clear and reliable recommendations for AI tools tailored to their specific needs. A copilot could serve as a more user-friendly and effective solution providing tool suggestions through messages are generated based on the latest information online.
I explored the layout and the balance between chatting and self-browsing experiences. I'm uncertain in terms of the product structure: whether to make the entire product a copilot or to position the copilot as a side assistant. Below, I delve into both concepts.
Copilot as the main interface:


After reviewing five people who are AI tool users with the interface above, I learned that
-
People could agree that AI chatbot can be useful for tool recommendations, especially for initial research or idea generation
-
People would still prefer to verify recommendations from other tool users, experts or through their own research before making decisions about AI tools.
Copilot as the side assistant interface:


In this version, I positioned the copilot on the side, allowing it to be toggled in and out. The main interface displays a detailed page of the tool, featuring collected projects, tutorials, and forums that help learn more about it similar to iteration 1.
I also reviewed this version with the same five testers and received more positive feedback, noting that it facilitates self-research for verification. However, there are still questions from both myself and the testers:
- The copilot and tool interface are separate - “I wish my chat and the tool information could interact with each other.”
- How can I encourage users to engage in self-searching and chatting to streamline the process of receiving recommendations and starting to use the tool through design?
Inspirations from browser tabs & Design Debrief
I began by reflecting on our everyday experience of learning a new tool or object. We explored various websites, often with multiple tabs open. At times, we open a new tab from one site or switch between tabs—this could be a way to integrate Copilot into the search and learning process to connect the copilot and tool information!
Design Debrief
Tab Experience Design

Example Chat

Connections of Copolit and Tool Page: Add To Chat

Split Screen

Other Features
Chat History for Revisit

Other Features
Discover Page for Latest News and Products

Other Features
My Saved Page for Revisit of Tools and resources

Results and Next Steps
As I reviewed this version of the design with the six people:
6/6 of them believe CosmoS is an innovative and intuitive interface
6/6 of them believe CosmoS is more effective in tool recommendation than general search on the internet
5/6 of them believe CosmoS makes the process from tool selection to starting using the tool more efficient
5/6 of them believe CosmoS gives more exposure to qualified AI Tools
4/6 of them believe CosmoS would help effectively them find the right AI tool
3/6 of them believe CosmoS reduced the learning cost of new tools
I found great satisfaction in reimaginingilot experiences and developing effective, innovative solutions. Based on the review feedback, I aim to incorporate more design details and concentrate on minimizing the learning curve. One suggestion I received was to label the learning cost on the tool card or include filters. Additionally, I can also add labels for features like affordability to assist users in choosing the tool more easily, rather than having to type manually each.


