ClearSight AI
Developed an innovative AI framework that enhances user experience by prioritizing company privacy and data accessibility for business owners. This approach aims to cultivate transparency in the AI system, providing users with a clearer understanding of its outcomes.
My Role
UX Designer
UX Researcher
Team Members
Seowon Yang
Carl Tischer
Bill Torson
Duration
4 weeks
Tools Used
Figma
Miro
Blender
Target Audience
Business Owners
How it started?
This project started from a conversation that our team had. Our team is a combination of UX Design, Industrial Design, and Business Management major. One of our teammate who works for a AI company, provided us an insight that it is not easy to convince business owner to utilize AI for their business. Our team saw this as an interesting topic that we did some research.
During our research, our team figured that in the past decade, there has been a noticeable surge in using artificial intelligence(AI) within the business world. Although business interest in AI continues to expand, there remains a hesitancy among business owners to utilize AI for their business. This reluctance stems from the absence of legal regulations safeguarding companies post-implementation. Our team saw an opportunity in this situation that we chose to proceed this topic as a project.
Pre-project research was conduced in order to allow our teammates to deeply understand the problem and an opportunity to fill the gap between business owners and AI was found. Our team challenged ourselves with the solutions for the problem that business owners were facing with utilizing AI.
Pre-Project Research
Before going further with this project, our team wanted to understand the situation in more detail that we conducted a pre-project research by going over AI Flow Chart and organizing user interview based on the information we gathered from relevant articles.
AI Flow Chart
In order to understand how AI generate answers, our team went over the AI flow chart. This was an important step for us since some of our teammate were not AI experts.The flow chart of how Generative AI processes information and creates responds is shown below.
Relevant Articles
Our team gathered relevant articles to understand the current situation between AI and business owners.
By going through this flow chart, our team was not only able to understand the flow of the AI but also saw some questionable area where business owners might be facing difficulty with. Our teammates notes are shown below.
User Interviews
In order to understand specific concerns the business owners have, our team conducted user interview. With one of teammate’s connection, our team was able to gathered participants from business owners of Fortune 500 who are also potential user for this project. Since this was a pre-project research, I created simple questions that will help our team to assess the business owners familiarity and concern about using Artificial Intelligence(AI). The questions are listed below.
“What are your duties and responsibilities?”
“Have you used AIs like Chat GPT?”
“How has AI assisted you in performing your duties?”
“What potential do you see in using AIs like Chat GPT?”
“What concerns do you have about using AIs like Chat GPT?”
Findings from User Interviews
Thankfully, our participants responded each questions with thorough response and these were their final sentence of the user interview.
Katie (VC Startup Headhunter, Female)
“Generative AI is amazing but don’t let it write your VC pitch and don’t lose your IP”
Tim (Investor and Researcher, Male)
“The markets are demanding regulation of Generative AI.”
Earl (High-Performance Computing Expert, Male)
“The rise of Generative AI is the next step beyond deep learning and part of normalizing AI in everyday life.”
By going through User Interview, our team gathered major opinions of the business owners have toward AI.
AI could assist greatly with productivity in the future.
AI could jeopardize IP protection and copyright.
AI is not always accurate in its output.
There is not a lot of trust in AI with user’s data.
AI poses many legal risks.
The research provided a clear narrative the business owners have towards AI and also, an understanding of what makes them resistant to use AI even though they think AI is a productive resource.
Problem
During the pre-project research stage, similar opinions overlapped each other and concluded into these kind of problems:
Users have no idea as to where the AI gathered its information, how it processed that information, or how it assembled the response.
There is uncertainty as to who owns the outputs of the AI itself. There is no legal framework for an AI to own anything.
Sending confidential information to AI will compromise that information. The AI can unknowingly use copyrighted work.
We categorized these problems into three sections: No Transparency, No Ownership, and No Protection and Confidential.
Solution
After brainstorming ideas, our team concluded with a solution of creating an AI that is more transparent with its result which allows business owners to feel safe to use by protecting the company privacy and data.
1
Our Mission
2
Brainstorm Ideas
3
Prototyping
“How might we protect company privacy and data for business owners using AI, while making the AI more transparent about its results?”
— Our Mission
Brainstorming Ideas
Based on the problem resulted from the pre-research, our team generated ideas for each problems:
No Transparency, No Clear Ownership, No Protection and Confidential.
For No Transparency:
Provide a visual representation of how the document is created.
Make AI provide a highlight where the source is used on the document.
Utilize clear and simple colors that give more transparent feeling to the users.
Create a function that lets the users filter some sources out of the information.
Make AI cite its sources.
For No Ownership:
The owner who has the authority to the source will be asked whether they approve someone else’s usage of the data.
Whenever someone utilizes a source, an alert would be sent to both the owner and the users.
Each source will be labeled with owners name and their role of the source.
Each alteration to the source will be recorded to the history.
Showcases the funnel of the source.
For No Protection and Confidential:
AI source information is localized on-site.
All the information the AI sources is labeled.
AI shows where the output is coming from.
Users are limited in what information they can access.
Users must audit the outputs to identify problems.
Source is segmented into different levels of access.
Selecting Concepts
We evaluated the brainstormed ideas by dividing them into different segments of the user experience and grouping them together. A few themes emerged from our findings that overlapped across the entire experience.
After reviewing the themes, our team merged the generated ideas into three major concepts:
Flow Visualization
Labeling Sources
Layering Security Access
Our team generated concepts for each problems that generative AI has; however, it was to vary to explain to our users that we when through an solution aggregation session and concluded with creating an interactive AI which is a visual representation of our solutions.
Features & Functions for our AI
Based on the selected concept from the brainstormed ideas, our team proceed another brainstorming session to provide more specific features and functions that allows users to view the visual representation of our solutions. The features and functions for this AI will be:
A panel that shows where each part of the document that AI creates is sourced from.
Allowing users to check the input and output sources and the Cloud LLM.
The users can return and ask the AI to generate a new document.
The users are able to check the history and see the previous work that AI-generated.
Before the company data is utilized for AI, a warning pop-up will be asked whether the user wants the data to be private or public.
Each resource will be labeled with the owner’s information.
Some resources will have limited access if the user has no ownership towards the resource.
A unique system of user’s input data will be secured even when it is used in AI
System Diagram for our new AI
Our AI allows users to combine the results of two different queries to create new, trusted generative content. Since the system provides a transparency with the system and the result that the name of the program is “ClearSight AI.”
More specific instruction for ClearSight AI is written down below:
User prompts ClearSight AI to search through data and pull significant insights into a formatted output.
User prompts ClearSight AI to collect a non-specific content output from an external generative AI without disclosing any private information.
User uses ClearSight AI to combine internally data-driven insights with the non-specific content output to produce a final result.
What is LLM?*
For example: User typed "Selling a surfboard to Chad” to ClearSight AI.
Internal query discovers Chad’s interests, payment history, and profile from proprietary data
External query uses Chat GPT to create a great ad for a surfboard
ClearSight AI combines the two datasets to create a high-quality, personalized ad for Chad without compromising company data
Prototyping
Due to time constrain, our team made a decision to focus on the resource checklist page rather than design the whole visual system of ClearSight AI.
Rough Sketch
As a UX designer, I created a “resource check page” rough sketch during our team meeting based on the suggestions that was discussed on Brainstorming Ideas stage. In order to provide an understanding to those members who are not in design major, I added text detail on each elements.
Medium Fidelity Design
Low Fidelity Wireframe
To provide a clear understanding for our industrial designer, I re-created the wireframe with Figma and sent it.
Based on the sketched wireframe that I created, our teammate Carl who is an Industrial Designer made a visual representation with blender.
Lesson Learned
The scale of this project was much larger than what it was planned that it was difficult to create a significant result in a time limitation. In the future projects, I would like to plan out the project thoroughly in detail that the project is able to finish it on time.
Team project is always a difficult project since not everyone is available on the time. However, if one leader manages to organize the group and keep in touch with every member the difficulty will reduce significantly that I am glad I did not give up hope to our team members and updated every process of the project via email with the others who were unavailable for some meetings.
What’s Next?
I would love to take this project further and develop a more detailed wireframe with more pages to create an improved interactive high-fidelity prototype so that in-person or zoom user testing is possible. Also, brand image and logo design is not specific yet that I would like to do a branding for this AI project.