I led the concept, research, and design of CaseMatters Evo — an AI-native case management system built to replace Access Legal's flagship desktop product. The concept secured £20M in investment. We delivered H1 in eight months, against a three-year original plan.
Access Legal's Proclaim has served UK law firms for over three decades. It's deeply embedded across conveyancing, litigation, personal injury, family law, and probate. But it's a desktop-era application running on Citrix, with navigation-heavy interfaces that force users through multiple screens for tasks they perform dozens of times a day.
During stakeholder interviews at JMW Solicitors, a senior fee earner gave me the problem statement that defined the entire programme:
“Often a 1-second job can take up to 15 minutes just trying to access it in the first place.”
The Access Group runs an internal programme called Maximus, where any software company within the group can pitch concepts to accelerate growth. Access Legal’s senior leadership — the General Manager, Divisional Managing Director, and Product & Engineering Director — asked me to conceive a completely new legal CMS. No legacy constraints. No inherited technical stack. Fresh thinking for how law firms work today.
I led the concept and design within a small team. I created the user journeys, prototyped the key interactions, and produced the pitch deck and video that brought the vision to life. Multiple software companies across the group submitted entries — some with multiple concepts. Senior leadership presented our concept to the CEO and board. They selected it for investment.
The concept secured £20M in investment and became CaseMatters Evo — Access Legal’s next-generation AI-powered case management platform. Originally scoped as a three-year delivery programme, I helped compress H1 to eight months by integrating AI into the design and build process.
I designed and ran a multi-stream research programme to make sure every design decision could trace back to real evidence rather than assumption. The goal wasn’t decorative research — it was a decision-making engine I could point at throughout design and delivery.
A quantitative UserZoom survey reached 110+ UK legal professionals across roles and practice areas, producing priority rankings, workflow patterns, and pain point data. In-person stakeholder sessions at JMW Solicitors gave me the qualitative depth — the “15-minute access problem,” the approval workflow pain, the email-first morning routine. Task-based usability testing then validated every major design decision before it reached development.
Later, I ran a six-firm cross-firm research programme — Bond Turner, JMW, Minster Law, MAPD, Fletchers, JF Law — generating over 30 hours of recorded feedback from 40+ participants, validating the platform against real working environments before wider release.
Fee earners had no single source of truth. Email, Proclaim, billing systems, and external portals all held different fragments of the same case. This drove the design of a unified case timeline as a core architectural decision.
Court deadlines, exchange dates, limitation periods. Missing them has professional and regulatory consequences. This shaped the visual hierarchy of every screen in the system — urgency had to be immediately readable, not buried.
Fee earners began every day in Outlook, not Proclaim. Rather than fight that habit, I designed email integration as a day-one feature — the platform needed to meet users where they actually started.
Testing repeatedly showed that card-only designs with minimal metadata felt insufficient. Users needed case references, client names, dates, financial figures, and status indicators simultaneously — information density was a feature, not a problem to solve away.
Moderate trust for routine admin; low trust for high-stakes decisions. This became the governing principle for every AI feature: surface intelligence and show reasoning, never act autonomously.
“If you change it and it’s as slow as it is now, it will go down like a lead balloon.” — Rachel Wong, Director, Bond Turner. Speed wasn’t a nice-to-have. It was the threshold for adoption.
AI as recommender, not decision-maker. Every AI feature surfaces intelligence and shows its reasoning. None act autonomously. Users can dismiss or override with one action.
CaseMatters Evo spans six interconnected modules. Rather than walk through each one, these are the four design decisions that had the biggest impact — the moments where my judgment shaped the direction of the product.
Decision-first workflows over task lists
Competitive analysis of 100+ legal tech platforms showed every competitor presented task lists. None had solved decision-first workflows — surfacing what decision is needed, why it matters, and what happens if you delay. I designed the workflow interface to present context alongside actions: not “Complete task: Obtain redemption figure” but “Confirm redemption amount from lender — Required for SDLT calculation — Exchange deadline: 12 Nov.” This approach was architecturally enabled by Azure Durable Functions, which can pause a workflow mid-execution at near-zero cost during dormancy — dropping workflow state management costs from an estimated £1,200–1,500/month to approximately £30/month for a 30-user firm.
The Action Desk: replacing navigation with proactive intelligence
Fee earners had no single starting point for their day. They opened Proclaim, navigated to individual cases, checked emails separately, and maintained mental priority lists across multiple systems. I designed the Action Desk as a Kanban-based workspace that surfaces everything requiring attention — across Overdue, Today, Tomorrow, and This Week — without the user needing to go looking. Each card follows a consistent five-zone anatomy (Z1–Z5) with fixed placement of trigger context, metadata, urgency indicators, and primary CTA. I mapped 13 card types across three tiers: day-one integrations, Proclaim parity, and net-new capabilities.
A 25-status system that doesn’t overwhelm
The legal accounts ledger is one of the most complex areas of legal software — managing client and office account transactions under strict SRA compliance. I designed a 25-type status indicator system spanning transaction, ageing, client, matter, posting, workflow, and compliance statuses. Each renders as a single word plus colour, replacing the icon-heavy approach of earlier iterations. The key trade-off was between completeness and cognitive load. I introduced a “Clean View” that hides reversed transactions by default — giving fee earners a clear summary — with a toggle to expose the full audit trail for cashiers. Two roles, competing information needs, one interface.
AI reasoning over time-saving badges
Early designs showed “Save 5 mins” badges on AI-prioritised cards. Testing showed this was too abstract — fee earners think in actions and consequences, not time savings. I changed the design to contextual reasoning tags: “Court deadline tomorrow,” “Client awaiting response since Tuesday,” “Settlement authority expiring.” These performed significantly better because they answered “why should I care?” rather than “how long will this take?”
CaseMatters Evo was originally scoped as a three-year delivery programme. I helped deliver H1 in eight months. The difference was integrating AI into every stage of design and build — not as a shortcut, but as a force multiplier that let a lean team operate at the throughput of a much larger one.
I used AI as a persistent design partner throughout the project. Research synthesis that would traditionally take two to three weeks completed in two to three days — not by summarising, but by mapping every data point exhaustively against design decisions. Competitive analysis across 100+ platforms was systematically structured rather than sampled. Specification documents were generated directly from validated design decisions rather than reconstructed after the fact.
The critical principle throughout: AI augmented my judgment. It never replaced it. Every output was validated against research, challenged against Proclaim’s data model, and iterated based on stakeholder feedback. The speed came from eliminating administrative overhead, not from bypassing design thinking.
Using Evo Builder — Access’s AI-powered development platform — I translated research-validated UX designs into structured prompts that generated production-quality React components using the EVO Design System. A component that might take two weeks in a traditional designer-developer handoff cycle reached production quality within days. The workflow was systematic: design through research and wireframing, translate to a structured prompt, review the rendered output, iterate through targeted corrections. Every change committed to a code repository — traceable and reversible.
A concept pitched to the board became a £20M investment and a product that compresses legal workflows that used to take 15 minutes into seconds.
Between November 2025 and January 2026, I took CaseMatters Evo to six UK law firms for on-site research sessions. 40+ participants across IT Directors, Partners, Fee Earners, Legal Secretaries, Paralegals, and compliance specialists gave us direct, unfiltered feedback on the product in their own working environments.
The AI capabilities, document bundler, and Outlook integration generated the strongest positive reactions. JMW rated the platform 9/10. But I designed the sessions to surface problems, not just gather praise — and they did. Timeline density needed to increase for volume PI caseloads. Performance remained the single biggest adoption risk. Screen layouts needed to respect existing muscle memory.
“If you change it and it’s as slow as it is now, it will go down like a lead balloon.”
I categorised every finding using a priority framework — Critical, Quick Win, Roadmap, Parking Lot — and fed them directly back into the design. This wasn’t a validation exercise to confirm what we’d built. It was a course-correction engine.
This wasn’t a solo effort. The product was shaped through close collaboration with product management, engineering, and external stakeholders — each bringing constraints that made the design sharper.
I established a clear working boundary with the PRD from the start: it defines functional scope, not layout. That meant I could propose different information architecture, interaction patterns, and card layouts while staying within agreed feature scope. Where I found gaps — missing reject flows for slip approvals, incorrect CTA labels, a 50-card cap with no UI representation — I raised them before they reached development.
Engineering collaboration was essential for a product bridging legacy and modern infrastructure. I checked every feature I designed against Proclaim’s data model via the CDM Bridge team. This discipline prevented the expensive mistake of designing interfaces for data that doesn’t exist — and it caught issues early. Features that looked compelling in competitive analysis but had no Proclaim equivalent were systematically removed from MVP scope rather than built speculatively.
Designing at the boundary of legacy and modern infrastructure means constraints arrive from both directions at once. The best design decisions in this project came from understanding both sides well enough to know when to push back and when to adapt.
Some iteration cycles were spent on features that later had data model gaps. I caught most of them — but catching them at the sketch stage rather than the prototype stage would have saved time. Next time I’d establish the CDM Bridge relationship at the start of concepting, not after initial wireframes were in review.
The UserZoom surveys and usability tests captured moments in time. Watching how fee earners actually move through their entire day over a week — especially the transitions between desk work, court, and client contact — would have given richer workflow data. Snapshot research answered the right questions; it just couldn’t show us the patterns between them.
Several patterns I created for this project — the card anatomy, the 25-type status system, the row interaction patterns — could have been contributed back to EVO as reusable components from the start. Setting up that feedback loop early would have benefited other Access products sooner, and given the patterns more rigour through wider use.