Relativity

Designing new features that speed-up tedious & time-demanding document review for legal teams.

Relativity is a legal technology company known for its RelativityOne platform, a comprehensive eDiscovery and legal case management system used by law firms, corporations, and government agencies.
I worked on a product called aiR for Review, which uses gen-AI to review the relevance of individual documents in a large document set for a court case (which typically contains thousands of documents).

I designed a new feature called aiR Insights, which surfaced key information on individual documents such as document summaries and red flags. This helps speed up document review, saving on client money and manpower.

Product Design Intern

Product Design Intern [Elaine Guan]
Product Manager (x2)
Applied Science (x2)
Software Engineering (x3)

UX Design
User Testing, User Research
Prototyping
Content Design

Figma, Figjam

Document review is a complicated and tedious task. Currently, aiR for Review users also have low visibility to document content when using the product dashboard, leading to a more time-consuming workflow.
‍
Some have turned to other tools as a result, and Relativity wants to address this gap and provide more value to customers through a built-in insights generation feature.

How might we empower legal teams to uncover the most critical information in documents, so they can review faster, reduce risk, and deliver with confidence?

Overview

Problem

Solution

Context

Process

Research

Lo-fi

Testing Insights

Refinement

Impact

Learnings

Surfacing key information and insights for reviewers to reference easily in document review.

This feature, aiR Insights, uses AI to surface key information, summaries, and contextual insights from documents, accelerating review and improving quality control. In this video of the a4R dashboard, aiR Insights is integrated into the overall a4R workflow.

aiR Insights saves client time and money by surfacing information that is pivotal to legal cases.

Identifying red flags & critical content ASAP.

aiR Insights flags and surfaces critical content in a document set.

90% of litigated cases* are settled before going to trial, often because a party could have documents with private or embarrassing information that they don’t want to publicly disclose.

Identifying critical content that would lead to settlement as early as possible is essential for saving a legal team’s time and money.

*Litigation means the process of taking legal action

a4r dashboard with air insights results

examples of red flags

a4r dashboard with air insights results

example of content summary

High-level content overviews that speed client workflow & improve internal communication.

aiR Insights generates 1-2 sentence long summaries that provide a high-level description of the document content.

Document summarization helps give context to relevancy results (an existing a4R feature), empowering reviewers to make more confident decisions. This information is also often used for internal team or client communication.

One client reported a 50% increase in user workflow speed.

A comprehensive product for a complicated industry.

Relativity is a legal technology company known for its RelativityOne platform, a comprehensive eDiscovery and legal case management system used by law firms, corporations, and government agencies. Through Relativity, clients can sift through massive volumes of data, analyze it, whether it’s for litigation, internal investigations, or other legal matters by leveraging AI to speed up and automate tasks & identify critical information.

During litigation (the process of going through a lawsuit), Relativity is used in the Discovery phase.

The feature I worked on, aiR Insights, is a part of the aiR for Review product, which uses LLMs to review and find relevant documents to a particular case. It identifies relevant documents and describes why they are relevant while using citations from the document.

Designing collaboratively with internal teams & external stakeholders.

Throughout the aiR Insights project, I collaborated with different roles and teams to design a product that balanced client value and internal company priorities. Specifically, I:

Collaborated with product managers to design the most optimal product UX given the context of the Relativity platform and long-term business goals.
Looped in engineers throughout the whole design process to ensure that our designs could be efficiently implemented.
Worked closely with Applied Science (prompt engineering) to create the categories and outputs that the feature would provide, balancing feasibility and utility for clients throughout this process.
Met with internal & external stakeholders and utilized feedback to ensure that we were designing something that fit seamlessly into existing customer workflow.

Understanding current client pain points & competitor offering

We conducted a competitive analysis of different e-Discovery softwares and also synthesized client feedback.

Customers need varying types of document grouping & summarization.

Through this, we understood that clients needed more context (in the form of categorization & document information) for their documents, and were starting to use different external tools (such as Copilot) to get this information. We organized the types of information clients needed into 4 broader content categories:

Topical Analysis, Categorization

Key document identification

Mapping themes & trends

Broad summaries

Different explorations for generating & displaying Insights results.

In our lo-fi prototyping, we explored different flows & placements of aiR Insights. For example, we explored keeping it separate from the a4R workflow vs. with the a4R workflow.

Breaking down the why behind failures in early prototyping.

Early Lofi Prototype – Distinctly seperating aiR Insights functionality from a4R.

This was an early lo-fi prototype that we presented to the team. At first, we believed that separating aiR Insights from a4R made more logical sense, but feedback from team presentations led to us pivoting from this.

There are some learning highlights from this prototype that we applied in future designs.

Balancing maximizing feature functionality with achieving design goals.

Part of aiR Insight’s value proposition was that reviewers receive results out-of-box (without prompting or custom input). This way, the feature can inform prompt criteria writing.

Needing to input custom categories defeated this purpose.

Creating customizable useful features is important, but aligning them with design goals and keeping in mind client needs is just as important.

We removed custom inputs and conducted desk research on litigation terms & processes to decide what the best outputs for categorization would be. Then, we validated the outputs via user testing.

Accounting for all design constraints and looping in engineers into the design process as early as possible.

aiR Insights was set to release in September and design was scheduled to be finished in July.

There wasn’t enough time & engineering manpower to code custom elements like these buttons & toggles. Also, the filtering controls & table were built in component library items and were hard to add/remove elements from.

To collaborate closely with engineering as early in the process as possible. In this case, that meant designing within & as closely to the design system as possible to ensure the design is feasible.

We moved the aiR Insights toggles to places where engineers could easily add them, and made sure to use out-of-box component library elements like toggles and tabs.

Designing with other interactions within the product is essential when trying to simplify client workflows.

After clicking on the control number of a document, a document viewer with more details on the a4r job would open.

The flow of opening another tab with aiR Insight results in detail was too similar to opening the document viewer, and would create friction when clients wanted to view both results in depth.

Accounting for other interactions within the product is essential when trying to simplify client workflows.

We integrated aiR Insights results into the document viewer in further prototyping.

Narrowing down features based off clarity and feasibility.

During team discussions, we decided on 3 main changes to implement in future prototyping.

Integrating aiR Insights into the aiR for Review workflow to ensure cohesion with the existing product.

Using design visual elements and solutions within the component library to minimize custom engineering work.

Keeping the aiR Insights results directly next to a4r results to maximize client value of aiR Insights.

Aligning with business goals & long-term product planning.

Following mid-fi prototyping, we began interviewing with internal and external stakeholders. We received feedback from a range internal teams like Product Marketing, Applied Science, and Extended Sales. While conducting external testing, we also interviewed with various types of clients, such as law firms, corporations, and legal service providers (LSPs).

From testing, we summarized four insights to continue with design refinement.

Greater intentionality with wording (with the context of the greater product suite) ensures user clarity.

Contextualizing AI-generated content with grounded evidence from individual documents best supports reviewer understanding and confidence.

Balancing the level of specificity aiR Insights provides is essential – surfacing too much information can cause greater confusion.

Insights bring the most value when paired with other aiR for Review features, not when it’s used exclusively.

Refining naming & feature outputs to clarify utility and match client expectations.

Rewriting names and descriptions to succinctly describe the feature.

During testing, a recurring theme that came up were clarifying qusetions or confusion with aiR Insight feature naming from clients. So, we started brainstorming different names to features to prevent confusion, especially with other RelativityOne features.

An example of this in the "Content Type" feature

Narrowing down output specificity to maximize value and minimize information overload.

Through testing, we found that different types of content are useful at different levels of specificity. So, when deciding the outputs aiR Insights would produce, I brainstormed and categorized the individual outputs with other team members. Then, we decided the level of specificity that would be shown to the user.

Accounting for edge cases in product usage.

When the Agile sprint for development of aiR Insights started, we conducted a brainstorming session to see what additional tasks designers and PMs had to complete. During this session, we identified different edge cases where Insights might not run or where a specific toggle or document might error. We worked with a PM to ensure that we accounted for and designed for these different edge cases.

Designing for long term growth and scalability.

In early stage user testing, we received feedback from different clients on how aiR Insights contained highly requested functionality that they had wanted for a long time, and was a tool that easily fit into and improved current workflows.
Presentations internally were also well received and met with support.

Relativity's long term goal is to implement aiR Insights into it's other product offerings.

The work created for aiR Insights in aiR for Review was the first step & pilot test for this long term vision.

Experimenting with and learning about emerging technologies.

Understanding why the product works.

Close collaboration with applied science and engineers gave me a close look into how labelling datasets and prompt engineering functioned behind the UI.

Designing with emerging technologies.

Experimenting with Figma Make and various AI design-to-code tools for early stage prototyping and concept testing helped me grasp how these tools work & where they enhance the design process.

Empathizing with unique clients.

Conducting and sitting in on user interviews allowed me to meet the wide range of clients Relativity has, and how even similar clients (ex. mid-sized legal firms) can have completely different workflows for the same tasks.

Utilizing novel testing methods.

I facilitated a early concept testing call that utilized Figma Make. Seeing this interactive prototype in action really excited me about how future user testing could feel more immersive and yield more insightful results!

Close cross-disciplinary collaboration.

Since the value of certain features was so tied to what was possible from a development end, close collaboration and discussion on how features worked was crucial to project success.

Learning to design with developers.

Keeping in mind developer contraints and priorities in sprints helped me create the most feasible and effective designs. I learned to loop in developers earlier in the design process and to work with them throughout the process.

I really enjoyed this internship and learned so much working in legal technology, a space I had no experience with before. I came out of this internship with...

...experience on how to effectively collaborate with cross-disciplinary teams
...understanding the process behind how prompt engineering and labelling data for AI training works
...a greater understanding of the intricacies & jargon in the legal industry
... & my first visit to Chicago!

Thanks for viewing my work!

aiR Insights was one of many projects I worked on over the course of my internship. If you're interested in learning about my other work for Relativity, feel free to reach out to me!
‍
Some other tasks I worked on were:

Redesigning the flow for creating new versions of a prompt criteria in a4R
Creating responsive designs for the dashboard for accessibility purposes
Designing and refining UI for a new aiR for Case Strategy feature
Various UI audits across a4R

Linkedin↗