How might we make investment documents easier to analyze for everybody?
Interface design for an AI-powered investor analysis tool
Time
Concept: 2 weeks, group
Design: 1 week, solo
Design: 1 week, solo
Class
Design of AI Products and Services (05-317)
User problem
Investment documents are hard to read.
Investor relations documents can surpass 75,000 words, yet contain valuable kernels of investment information. Large investment firms have the resources to comb through these documents, but not the average personal investor. How might we make investment documents easier to analyze for everybody?
InvestEdge breaks down and analyses Investor Relations documents. To do this, InvestEdge uses Large Language Models and other NLP algorithms that are capable of summarization, sentiment extraction, and abstract problem-solving to help guide investors with their portfolios.
InvestEdge breaks down and analyses Investor Relations documents. To do this, InvestEdge uses Large Language Models and other NLP algorithms that are capable of summarization, sentiment extraction, and abstract problem-solving to help guide investors with their portfolios.
Defining users & strategy
Primary user
- Mass-market or affluent investor (stock portfolio >$150)
- Tends to make specific stock/security picks, rather than rely solely on ETFs
- Executes trades self-directed via brokerage, but may have an FA available
Design considerations
- InvestEdge should not make choices and should only act as a tool for analysis.
- InvestEdge should be visually digestible.
- Users should be able to view both the big picture and concise insights.
- Because this is AI, InvestEdge should be designed to accommodate for failure cases.
A framework for visual & interaction design
Low-fidelity wireframe
Using a low-fidelity wireframe, I created a simple mobile app for InvestEdge.
![]()
Using a low-fidelity wireframe, I created a simple mobile app for InvestEdge.
Visual identity
I created an easy-to-digest and trustworthy visual language through a clear hierarchy and a spacious layout.
![]()
I created an easy-to-digest and trustworthy visual language through a clear hierarchy and a spacious layout.
Flow
InvestEdge preserves the content of the document, while prioritizing important content. It does this by highlighting important content with NLP, while deprioritizing content that’s less relevant.
Understanding Nvidia’s annual report in five minutes
InvestEdge preserves the content of the document, while prioritizing important content. It does this by highlighting important content with NLP, while deprioritizing content that’s less relevant.
Home page
Users see a list of documents ordered by date, with categorized text insights being highlighted.
Users see a list of documents ordered by date, with categorized text insights being highlighted.
Filing page
Once users tap into a document, they can read AI-picked highlights. They can also read skipped text by expanding it.
Once users tap into a document, they can read AI-picked highlights. They can also read skipped text by expanding it.
Audio mode
For people with vision impairment and for people who want to listen to content, pressing the play button narrates the text. An indicator follows along with the narration.
For people with vision impairment and for people who want to listen to content, pressing the play button narrates the text. An indicator follows along with the narration.
Scrubbing
By moving the scroll bar, users scroll through different sections. They also can preview content, which is represented by NLP-summarized headlines.
By moving the scroll bar, users scroll through different sections. They also can preview content, which is represented by NLP-summarized headlines.
Flow
Onboarding, picking companies and priorities
Sign-up page
Users can authenticate with Google/Apple or create an account.
Users can authenticate with Google/Apple or create an account.
Adding companies
Users search and add publicly-traded companies to their watchlist.
Users search and add publicly-traded companies to their watchlist.
Ordering categories
Users drag and drop categories to rank their investment priorities.
Users drag and drop categories to rank their investment priorities.
Home screen
Users are then brought to their home feed.
Users are then brought to their home feed.
Closing thoughts
My favorite part was...
The concept of matchmaking– basically, you go over a list of AI capabilities and match it to user needs in a certain area. Through this, there are three things that were particularly relevant: summarization, extraction of meaningful data, and the contextual answering of user-generated queries. This can then get connected to investor needs!
If I had more time...
I would explore the ability for LLMs to answer a user’s query about a body of information. This by itself is a project of it’s own, and I did not investigate this within InvestEdge due to it’s massive scope.
The concept of matchmaking– basically, you go over a list of AI capabilities and match it to user needs in a certain area. Through this, there are three things that were particularly relevant: summarization, extraction of meaningful data, and the contextual answering of user-generated queries. This can then get connected to investor needs!
If I had more time...
I would explore the ability for LLMs to answer a user’s query about a body of information. This by itself is a project of it’s own, and I did not investigate this within InvestEdge due to it’s massive scope.