Reimagining an AI Legal Research Platform for Better Discoverability and User Trust
Reduced task time by 35% | 80% fewer UI inconsistencies | 42% increase in AI feature adoption
My Role
As a product designer, I drove the end-to-end transformation of a feature-heavy AI platform into a clear, task-focused experience that improved discoverability, usability, trust, and adoption among lawyers.
Team
1 Product Manager
1 Developer
2 AI Engineers
1 Product Designer (me)
Skills
Visual Design
UX Design
Design system
Prototyping
Usability Testing
Timeline
5 months
Context
Legal Genius is an AI-powered research platform that helps lawyers analyze case files and generate drafts. As the Founding Product Designer, I led the 0-to-1 redesign of a platform originally built by engineering without UX foundations, high technical debt, and a 60% monthly churn rate.​
THE CHALLENGE:
Lawyers work under high pressure where every minute counts. Legal Genius was built to save time, but a confusing interface made basic tasks impossible. Users spent 30–40 minutes struggling to find value before giving up and returning to manual work.
OPPORTUNITY:
The legacy tool's friction created an opportunity to pivot from a complex technical product to a high-speed legal utility. The goal was to align the platform with a lawyer's need for instant value, verifiable AI accuracy, and increasing productivity.
CONSTRAINTS:
I was unable to conduct direct interviews with end users due to confidentiality policies. The tool has evolved based on client-requested features, so there was no prior research, analytics data, or UX foundation available.
Discovery & Research
01.
I collaborated with the customer-facing team to understand who our users are, what they are currently struggling with, what their workarounds are, and what their needs are.
"I didn’t know where to start. It was overwhelming and hard to understand. I couldn't figure out where to click with so many options, so I went back to doing it manually."
02.
Conducted a UX audit on the existing platform to assess accessibility, identify friction points and visual errors and determine areas for improving user experience, focusing on the key factors that influence it.
1. Decision Paralysis
Here’s a revised version: "Four buttons competing for attention; users had to think which one to click first."
2. Hidden AI Search
The main interface did not display the core AI search, making it challenging for users to locate it.
3. Technical Jargons
Due to technical jargon, such as "JSON file" and "query," the platform failed to build initial trust and familiarity.
4. No Guided Workflow
Upon signing up, users are taken directly to the home page without any guidance on what their first action should be.
5. Empty States Feel Broken
Empty states lacked value and contributed to visual clutter and confusion.
6. Long Forms Break Flow
Users had to fill out 9 fields for case creation and over 12 for additions, causing friction before realizing value.
03.
Mapped the standard 0 to 1 legal journey to identify how the tool can be incorporated into the existing framework.
This analysis helped me visualize the ideal user journey, showcasing where the tool can effortlessly integrate into a lawyer's workflow.
.png)
04.
To identify the product's shortcomings, I created a user journey map. This highlighted significant pain at each stage, ranging from confusion at the beginning to frustration and distrust with the outcomes. This confirmed the findings from the heuristic evaluation analysis.
.png)
04.
Our goal was to go beyond merely developing another AI tool and instead create a dependable legal partner. This guiding framework shaped our design choices, enabling us to focus on more than just basic features and to comprehend how the product addresses real-world challenges and operates in user scenarios.
Main Insights from Research
#1
For legal professionals, technical complexity is a liability; a tool is only valuable if it provides immediate, verifiable truth without a learning curve.
#2
Trust is the primary barrier to AI adoption in law; users will not rely on an automated answer unless they can instantly verify it against the original source document.
04.
From my research, I pinpointed the top three Jobs To Be Done (JTBD) that guided me in streamlining essential workflows and tailoring the user experience accordingly.
"When I am assigned a case with several documents, I want to extract a high-level summary, key points, and relevant judgements instantly so I can form a legal strategy without getting buried in paperwork."
"When preparing for a hearing or verifying opposing claims, I want to cross-reference multiple documents and see the exact source, so I can argue with 100% confidence and zero doubt."
"When managing multiple high-stakes cases, I want to organize and access key document details with the fewest clicks possible so I can stay in a flow state instead of fighting the software."
My Approach
I reframed the product around how lawyers already think and work.​​​ Instead of re-organizing the interface around features, I reorganized it around mental models and real workflows.
​Three principles guided the redesign:
-
Search should be the primary entry point
-
AI must be transparent and verifiable
-
Complex tasks should feel guided, not overwhelming
Every design decision was tied back to these principles.
FigmaMake to Concept Validation
1. AI Accelerated wireframing
To enhance the design process, I utilized FigmaMake to create and evaluate various concepts based on our research findings. This 'fail-fast' strategy enabled us to quickly validate ideas with stakeholders in real-time, reducing validation time by approximately 50%.
Within a few days, I developed three wireframe flows that, while not perfect, provided us with a clear direction.
-
Home
-
Case Management
-
AI Research Workflow
Home
Case Management
Stakeholder feedback showed the sidebar and AI search were promising, but too dense. I shifted focus to reduce cognitive load and enhance usability by simplifying case cards and using space effectively, making the tool more user-friendly.
02. Learning from Common Design Patterns
We met our users where they are rather than teaching new behaviors. By analyzing patterns from commonly used tools like ChatGPT and Claude.ai, we integrated familiar AI workflows that aligned with lawyers' mental models, reducing the learning curve and boosting tool adoption.
From AI Concepts to Intent Based Designs
01. Home: Search-Based Navigation
The previous dashboard featured numerous equally weighted actions, leading to decision paralysis.
Validation of the new concept showed that while navigation was simplified and information became more accessible in the "dashboard" style, the simultaneous presentation of all critical information still led to cognitive fatigue.
Old Home
Early Concept
Solution
Prioritizing the user interface around the main "AI Search" action rather than just data, we made it easier to begin a workflow.
-
Immediate task initiation
-
Reduced confusion around “where to start”
-
Improved feature discoverability
-
The platform shifted from “tool navigation” to “question-first thinking.”
02. Case Details
The earlier version depended on disjointed folders and disruptive modals, hindering the user experience and complicating high-level case overviews.
Old Case Management
Early Concept
Solution
I redesigned the modal-heavy structure into a focused case management hub. This centralized platform allows lawyers to effortlessly draft documents, request AI case information, and manage timelines and details all in one cohesive space while context.
​
-
Simple case organization and management
-
Immediate access to information related to cases and previous AI search and insights
03. Draft Creation
To create a draft with AI, users need to fill out a form. They struggled to find the "sweet spot" between manual drafting and AI generation. The lack of a clear workflow led to a steep learning curve and user frustration.
Old Draft Creation
Solution
I designed a split-interface approach that balances AI assistance with manual control. This allows users to seamlessly toggle between AI-generated suggestions and traditional editing, ensuring the tool feels like an assistant rather than a replacement.​​
Balancing AI Fluency with User Control & Transparancy
Lawyers were hesitant to rely on AI-generated responses due to the inability to verify the source of the information; blind trust is not acceptable in legal practice.
By enabling them to select the source of outputs, such as "Search Files Only," "Search," and "Search Past Precedents", confidence in AI-generated answers is improved. This fosters greater acceptance of AI recommendations and reduces the time needed for manual verification.
In the legal sector, "hallucinations" are a major concern. Users expressed discomfort when they couldn't trace the origin of AI-generated content. If users can't confirm the source, they won't trust the output, regardless of accuracy.
We enhanced the AI's output by presenting sources, relevant document excerpts, and associated judgments, offering lawyers the verifiable data necessary to ensure confidence in their work's accuracy.
This resulted in:
-
Increased trust in AI-generated responses
-
Higher adoption of AI suggestions
-
Reduced manual verification time
New User Journey of Legal Genius
Meet Meera - she is a 45-year-old brilliant solo lawyer, but she is constantly overwhelmed. Between managing 10+ active cases and reading through hundreds of pages. She needs a simple second brain to help her find the winning truth, past precedents, create effective drafts, and make the case stronger from all aspects.
01. Guided Onboarding
Meera visits the home page for the first time. Rather than feeling confused, is encouraged to explore a guided tour of the key features.
​
-
The Goal: To make sure she feels empowered and can using the platform right away without feeling overwhelmed by its functionalities and technology.
02. Easy Starting Point (File Upload)
Meera skips the formal case creation process and goes straight to uploading her case documents.
​
-
The Goal: To quickly get her started. Once the upload is complete, she can instantly interact with her data, reducing the time to reach first-time value.
03. Intent-Based AI Research (Query)
Meera has the option to select her preferred language and search source, which can include her personal files, the platform's judgment database, or the internet.
-
The Goal: This provides users with the flexibility and control over the appearance and origin of the information.
04. Co-pilot Draft Creation
Meera asks the AI to create a draft. Rather than generating it right away, the AI poses a clarifying question to identify the specific type of draft needed. After Meera provides a prompt, AI efficiently searches for relevant sources to produce a professional legal draft.
-
The Goal: This builds user trust in AI output by encouraging the right follow-up questions, helping user reflect on their desired outcome, and providing detailed information.
Meera can open the draft in edit mode, enabling her to either collaborate with AI or make manual adjustments.
The Goal: She maintains total control. She uses draft-specific AI features to cite sources, smart review the text, and strengthen her arguments with one click.
05. Case Management
Meera can easily access and manage all her cases in one place. She can access a full 360-degree view of any case, including its overview, documents, drafts, and past AI searches
-
The Goal: Assist users in efficiently managing their cases by transitioning from manual management to a unified platform where they can find everything and reference past cases in the future.
Scalable System to Bridge Design and Development
As part of the platform rebuild, I created a 0→1 design system to ground the product from the very beginning. I defined core foundations, spacing, typography, color, elevation, and semantic tokens early on to keep decisions consistent as the platform evolved.
I built components alongside real research and drafting workflows, making sure the system could handle dense legal content and dynamic AI outputs without losing clarity or trust. This helped reduce inconsistencies, simplified collaboration with engineering, and made development more efficient and predictable.
Reflection
What I Learned
-
Rebuilds are rare opportunities; getting the system right from the start saved us from future headaches.
-
Designing for AI showed me how to balance automation with human judgment, keeping users in control and building trust.
-
Intentional friction and clear cues can make complex workflows feel guided, not overwhelming. User control and transparency are crucial for building trust; users cannot depend on information if they are unaware of its source, particularly in sectors like law.
-
Strong foundations made everything easier — they didn’t limit creativity; they made the product feel clear, reliable, and scalable.
What I’d Do Next
-
Continue evolving the design system by adding advanced AI-specific components and patterns to support new workflows.
-
Test and refine interaction patterns based on real user feedback, focusing on trust, clarity, and human-AI collaboration.
-
Establish governance and documentation to ensure the system scales smoothly as new features and teams are added.