AI Prompted Case Builder
MY ROLE:
Principal UI/UX Designer, GenAI
MY TEAM:
Product Owner (x1)
UX Designer (x1)
Data Scientist (x2)
Director of Research (x1)
Full Stack Developer (x2)
UI Developers (x2)
Problem Statement
As a law practitioner, my research hours for a new case run longer than the actual billable amount of time. I rely heavily on available research tools to gather information, and turn to AI tools that are not legally sound to create long pieces of documentation, like legal briefs, memos, or even client communication.
It takes incredibly long and requires a complex system of tools to write 1 comprehensive legal document, and often takes hours to truly cross-check issues, citations, sources and typos, that it cuts into my time to build intelligent arguments to support my client.
Is there a tool that can make my research and drafting process simpler, so I can focus on building the arguments and asking the right questions to win my case?
Objectives
Reduce the amount of time spent on finding the right citations for a case
Talking to immigration, corporate and civil rights lawyers, we realized that a major chunk of their daily routine was spent in researching and analyzing large blocks of information relevant to a case, most likely other case documents to find relevant precedents. The tool needed to help them conduct this search & analysis in as little time as possible, using techniques like tagging and NLP.
Leverage semantic search principles (LLM, NLP models) to build suggestive and predictive interactions
The downfall of most AI-driven conversations is not that the AI needs more training, but the human needs more training in writing high-quality prompts. To help our users craft that high-quality prompt, the product had to include educational suggestions, and anticipate lawyers' follow-ups during the case building process.
Simplify the legal writing and verification processes
Due to the widely prevalent mistrust of AI tools for creating legal arguments, most lawyers spend countless time and effort verifying AI-written content with the actual case or law that is suggested as a potential source for an argument. The product, CaseBrainZ, would be exceptionally useful if it allowed users to verify original case details, summaries, connected cases and laws in close proximity to the AI-written responses.
Rapid Design Process
I used a combination of Agile and Rapid Prototyping processes to complete the end-to-end design of this proof of concept (POC) project in the first 2 weeks of August 2024. I then used the next 2-3 weeks to test, demo to users, gather stakeholder buy-in, gather investor buy-in and define iterations.
Just the process...
Observe
Define
Learn
Ideate
Prototype
Test/Feedback
Iterate
Personas
Based on my UX discovery surveys and interviews, I was able discern 1 major user and 2 supporting users who would likely use CaseBrainZ to facilitate their legal research.
Competitive Landscape
Legal AI is a relatively new field, but there are giants here that built AI services on top of their suite of legal research tools. The biggest competition faced by CaseBrainZ came from LexisNexis (Lexis+ AI), and Thomson Reuter's WestLaw (WestLaw Co-Counsel).
Differentiator(s):
This new product (internally titled CBZ+ Quill) had to be focused on simple, case-building processes, and offer constructive suggestions for "next steps" with the power of AI — features that were not offered directly by any other single competitor.
Product Inspiration
Demonstrable Prototype
The first version of the CaseBrainZ AI Prompt Tool was rapidly built over a course of 2 weeks. I met with the stakeholders every day to report progress, gather feedback and define iterations to feed a long list of requirements, most of which was written based on my competitive analysis.
The end product was a prototype of the minimum desirable product — with the features that, at the bare minimum, met user needs in a seamless fashion.
Design Testing for Iterative Improvements
The product had to be rapidly prototyped to be shown to potential investors, users and stakeholders. Since it's completion on August 19th, 2024, the master prototype has been showcased to over 5 Dallas-based law firms (including Fennemore Law, which had over 20 participants), multiple investors and I have led the contextual interviews, with over 8 user testing sessions with new users.
Results / Statistics
Insights from user feedback
Users found it easy to get started, and recognized the layout of the "answers" as similar to legal memo structures.
The split of case documentation into 4 main buckets made the research process more linear and easy to switch between tabs, to focus on growing/refining individual blocks.
Feature ranking by popularity
Users compared the interface to other Legal AI tools and missed the presence of a constant search bar to ask follow-up questions.
Some features requested by users, like case analytics and counsel statistics, tied into the design vision of the product and offered decisive direction in terms of iterations and next steps to scale the product.
Features requested
"I'm happy that the tool makes valid and sensible suggestions to build a solid case, even when I have insufficient information from my client."
— Parth, a corporate lawyer & legal mentor