Relativity
Relativity is a leading legal technology company whose flagship platform RelativityOne powers eDiscovery, investigations, and compliance for law firms, corporations, and government agencies.
This summer, I joined the aiR for Review team, which uses AI-powered review to help 300,000 annual users sift through massive volumes of data for litigation, investigations, and other legal matters.
Relativity’s aiR for Review workflow currently has no way to automatically summarize documents or notify users of red flags. Reviewers read line by line, increasing the risk of missing crucial information.
Often, clients realize too late that some documents contain sensitive information, causing cases to fall through, settlements to collapse, and Relativity to lose revenue from platform usage. This also led to some users to turn to competitors rather than use Relativity for their end-to-end case development.
I led the 0→1 design of aiR Insights in close collaboration with PM, UX research, and applied science (prompt engineering), delivering a feature that accelerates review by 50%, reduces cost, and lays the foundation for future AI capabilities at Relativity.
SOLUTION PREVIEW - MVP PROTOTYPE
THE PROCESS
02 CONCEPT TESTING & ITERATION WITH STAKEHOLDERS AND USERS
APPROACHING THE PROBLEM SPACE
After exploring multiple design options, we narrowed down to two lo-fi versions, weighing tradeoffs in scalability, use of space, accessibility, and technical complexity based on where aiR insights would be triggered.
Solution 1: Trigger in List
Solution 2: Located in Sidebar (Selected Solution)
Decision: We chose to have aiR Insights located in a the sidebar to ensure scalability for future capabilities, such as custom user prompting. Although Option 1 saves space, it’s more technically complex to implement and may not scale well with more advanced options.
After discussing early aiR Insights results prototypes with engineers, we found they introduced unnecessary complexity for the project timeline. We pivoted to explore two simplified patterns inspired by existing designs.
Solution 1: Toggle with Icon
Solution 2: Located in Separate Results Tab (Selected Solution)
Decision: We selected the separate tab layout to ensure aiR Insights remained highly visible. While the icon-based solution was more seamless, its limited size and placement made it less discoverable, and we prioritized visibility for this new feature.
As we finalized the MVP through stakeholder and customer feedback, we refined the categorization – merging overlapping areas, redefining unclear ones, and removing redundant sections. This process also allowed me to dive deeper into content design, revealing how even a single word can significantly shape user understanding.
Decision: 5 Red Flags, including Illegal Behavior, Reputational Risk, Inappropriate Relationships, Deception, Prompt Injection. 31 Content Types, including Operations & Compliance, Communications, Contracts & Legal Agreements, Corporate Governance & Strategy, etc.
Key learning: I initially assumed users would value more document information and customization, but testing revealed that additional options often created confusion rather than adding clarity or usefulness, reminding me once again that less is often more!
I facilitated workshops between UX, PM, and applied science (prompt engineers) to synthesize insights gained from user research and competitive analysis.
Through these sessions, we discovered key insights that shaped the MVP scope, including a breakthrough that several seemingly distinct categories had significant overlap.
Topics, themes, issues, and content had similar patterns, which led to conversations about customization and user prompting. These capabilities were out of scope for the MVP but slated for the next feature iteration.
03 EDGE CASE CONSIDERATIONS AND FINAL MVP
Working with engineering and applied science, we anticipated scenarios like failed Insights runs, unclear Red Flag citations, and varied summary format preferences. Together, we designed fallback states, clarified AI reasoning, and created adaptable UX patterns to keep the experience clear and trustworthy even when errors occurred.
04 IMPACT & FUTURE DIRECTION
I truly enjoyed my internship at Relativity and feel so grateful for the chance to work with such a talented and supportive team! From day one, I was welcomed into a fast-paced environment where I could dive into meaningful design projects and learn about legal technology, a field I had never explored before but quickly grew fascinated by. I was given significant autonomy and the opportunity to work on high-impact products and initiatives, which empowered me to take initiative and contribute proactively.
This experience strengthened my product sense, deepened my understanding of cross-functional collaboration, and gave me practical insight into working effectively with engineers, researchers, and product managers to bring complex AI features to life.
I quickly learned the legal review process and its technical constraints, which helped me identify where AI could add the most value. By validating assumptions early and iterating on feedback from PMs and stakeholders, I adapted designs to meet user needs and the MVP scope.
Synthesizes feedback from customer sessions with companies like Verizon and Morgan Lewis taught me to turn qualitative insights into product scope. Instead of a generic “summary” feature, we defined targeted use cases like content summaries for speed and red flag detection for risk.
Collaborating closely with engineers and prompt engineers taught me how UX decisions impact model performance. Understanding how AI interprets data helped me design more informed patterns and improved my ability to think like both a designer and a strategist!
During my time here, I had the opportunity to contribute to a variety of exciting projects besides aiR Insights, including:
Incorporating user feedback into a Validation metrics dashboard to reduce customer confusion and information requests
Creating responsive designs for the aiR for Case Strategy dashboard that prioritized accessibility
Designing and refining the UI for Relativity's new AI assist chat
Conducting UI audits across aiR products to improve consistency and usability
Exploring Chicago with my intern class and team!
Addresses critical customer need
Clients demand faster, smarter document understanding, and competitors are already delivering similar capabilities.
Drives tangible business value
Fills a strategic feature gap, boosting aiR for Review’s market competitiveness and overall product value.
Transforms user experience
Offers concise, actionable summaries that minimize fatigue, reduce errors, and dramatically accelerate review workflows.
Beginning with aiR for Review, Relativity plans to expand aiR Insights capabilities across its full product ecosystem. The designs and groundwork I created for aiR for Review served as the foundation and pilot for realizing this broader vision.
“Instead of reviewing word for word why a document was not responsive... we already just give them insights via our own system—an abstract of the document, a 6–8 sentence summary. My modeling is that 50% of docs can be decided on insights rather than the documents themselves. You can speed up manual doc review by 50% plus.”
“I want to promote aiR for Review across the industry. It’s good for the industry. Look how much more we can do with aiR and genAI… This would be huge for LDI (Legal Data Intelligence) use cases.”
