Skip to main content

Return to Project

Glad you made it here. Every phase had a business reason,
a deadline and a constraint. Here is how I balanced business goals, user needs and real constraints across 18 months.

DESIGN APPROACH

Incremental Revamp in 6 Phases

Phase 01 of 06

Research + UI Tweaks

Business Context & Priority
  • Client needed visible proof of value before committing to a full rebuild
  • New functionality was added within the existing product
  • No risky architecture changes
Design Thinking
  • Understand before designing anything
  • Used heuristic evaluation and usability testing to build an evidence base
  • Prioritise quick wins that build team trust for bigger work ahead
The Challenge
  • No design baseline existed
  • Team had vague tickets with no clear picture of what was broken or why
  • Any design decision without research would have been a guess
How I Solved It
  • Improved layout spacing and reading hierarchy
  • Fixed button visibility and labelling
  • Removed visual clutter competing with core content
  • Every change tied to a specific usability finding, not preference
Why moved phase 2 →The foundational work showed a deeper problem: search was broken at a structural level, not just visually
Phase 02 of 06

Search Functionality

Business Context & Priority
  • Legal firms bill by the hour, so search time is a direct cost to clients
  • Business identified search efficiency as their biggest retention lever
  • Time, business demand and engineering feasibility shaped every call made in this phase
Design Thinking
  • Search should work as a global command, not a dedicated section
  • Filters need to reflect how lawyers categorise documents
  • Validated design decisions through testing before bringing them to engineering, which meant fewer surprises and revision cycles later
The Challenge
  • Search was isolated on its own page
  • No way to cross-reference a search result with the document view
  • Filters buried and built around back-end logic, not how lawyers think
How I Solved It
  • Redesigned search so it is accessible from anywhere in the platform
  • Filters by keyword, tag, custodian and date without switching pages
  • Results structured so the most relevant documents surface first
  • File types got icons and logical groupings so emails and attachments were immediately distinguishable
  • Iterated on filter logic based on what usability testing showed before final handoff
→ Users could now find documents but reading them was still broken and frustrating. That became Phase 3.
Phase 03 of 06

Timeline + Document Reading Pane

Business Context & Priority
  • Users were leaving the platform mid-review and finishing work locally
  • Business saw this as a direct retention and adoption problem
  • We took an incremental approach here: ship a meaningfully better reading experience, learn from it, then iterate
Design Thinking
  • Be clear about what is a UX problem and what is a technical constraint, and treat them differently
  • Native view was a real user need so I advocated for it, but designed in layers so engineering could deliver progressively rather than waiting for a perfect version
The Challenge
  • Reading pane was too small and had no context
  • No way to understand how documents are related to each other chronologically, as they are listed as pages
  • The UX problems with the reading pane were clear and solvable
  • The harder challenge was that building a native document view was technically expensive on the frontend
How I Solved It
  • Timeline: minor layout and display adjustments, structural changes
  • Reading pane: rebuilt with adjustable layout, better document context and smoother navigation within documents
  • Worked with engineering on a phased delivery: core reading improvements first, enhancements layered in through later iterations as feedback came in
→ Reading solved, but team review had no structure at all and was a commercial blocker. That became Phase 4.
Phase 04 of 06

Document Review Experience

Business Context & Priority
  • This was built completely from scratch, no existing patterns to inherit
  • Enterprise clients were holding off on expanding licences until this was available
  • Rather than trying to design and ship everything at once, broke it into smaller chunks: manual assignment first, automatic assignment next, each informed by what the previous iteration taught us
Design Thinking
  • Every design decision was weighed against three things: time available, business demand and what was actually feasible to build
The Challenge
  • Designing a collaborative workflow from zero with no reference point
  • To reduce that risk, conducted a physical design walkthrough, stepping through each flow to identify cognitive load, catch gaps in the handoff logic and surface edge cases
  • Collected feedback, iterated, then brought it back into Figma
How I Solved It
  • Manual assignment shipped first: team leads can delegate documents with context attached
  • Automatic assignment came next, shaped by how manual assignment was actually being used in practice
  • Live progress tracking (dashboard), so leads have full visibility
  • Reviewers can annotate, tag, and through dashboard can have a complete view
  • Each iteration validated before the next chunk was started
→ Testing revealed lawyers were looking for relationships between documents, not just individual files. That insight shaped Phase 5.
Phase 05 of 06

Refining Review and Concepts

Business Context & Priority
  • Business wanted to position the product as AI intelligence, not just document management
  • The Concepts feature had to work for users doing real work
Design Thinking
  • Shift the product from document management to case intelligence
  • Flexible enough for different legal strategies without imposing structure
The Challenge
  • The interaction needed to feel natural to a legal professional while working within the constraints of what the model could actually deliver
  • Review refinements were also running in parallel so design capacity had to be split carefully across both
How I Solved It
  • Lawyers can group relevant documents by case context and share those concept groups with the team to speed up review.
  • AI does the heavy lifting, the interface stays simple. Legal professionals are not here to learn new technology.
  • Core grouping shipped first, AI-assisted clustering layered in after.
→ User-facing product was now strong. Internal teams were the remaining bottleneck. That became Phase 6.
Phase 06 of 06

Admin Interface and Upload System

Business Context & Priority
  • Every new case required manual support team intervention to set up
  • Key operations had no designed experience: upload flow, live extraction status, knowledge base creation, document summary and admin dashboard
  • Each of these was a handoff point where things broke, slowed down or needed someone to step in and fix it
Design Thinking
  • Started by understanding how internal teams were solving this manually before designing anything
  • Sat with the team to map how uploads and AI extraction actually worked in the background, so the interface could reflect the real process not a simplified version of it
  • Understanding the AI pipeline was essential: knowledge base creation, extraction status and summary generation all needed to be visible to users at the right moment, not hidden
The Challenge
  • Internal users had never been designed for, so there were no existing patterns or documentation to start from
  • The challenge was giving internal teams enough flexibility to handle complex case configurations while keeping the interface simple enough that no one needed training to use it
How I Solved It
  • Ran user interviews with internal teams to understand their actual workflow and what mattered most to them day to day
  • Mapped their current workarounds to identify where the biggest friction lived
  • Modular case configuration that internal teams can navigate without any help
  • Upload flow redesigned step by step with clear progress indicators and live extraction status so users always know what is happening in the background
  • Knowledge base creation and document summary given clear visibility within the flow
  • Permission settings labelled in plain language, no ambiguity
  • Shipped upload and permissions first, full configuration module second
  • Internal teams now onboard new cases without any support intervention