I redesigned how thousands of law firms track their serves.

We knew it needed work, but it was a big undertaking with lots of stakeholders and opinions. We agreed to take a phased approach, starting with one widget: chat.

Chat is a key differentiator. Clients can message process servers directly. But every message creates a review task for ops. The business goal was to reduce task volume. The design question: why are clients chatting so much?

I used Claude to classify 1,000+ chat threads by intent, identifying 16 distinct message categories. The pattern was clear…clients weren't chatting because they wanted to chat. They were asking the same question over and over: what's happening with my job?

The problem wasn't the chat experience. It was that we weren't telling clients what they needed to know.

“People are dealing with some of the most traumatic experiences they'll ever have. A divorce, a child custody dispute, somebody's taken off with the kid. The more information we can give them, the more reassuring it is to them that things are moving in the direction they need them to go.”

— Paralegal, mid-sized family law firm

When jobs get redispatched (natural disasters, server availability, access issues, client requests) our system created a brand new page. New job. New chat. New history. Clients with complex serves had 2–10 separate job pages to navigate. Which one was current? Where should they message?

That's not a product experience. That's a breakdown of trust.

“Jobs” were an internal concept that made sense for ops but meant nothing to clients. We needed a new mental model: one serve, one story, regardless of how many reassignments happened behind the scenes. The serve became the container.

Job tracker

Answers one question: is this progressing the way it should? Instead of binary status labels, it surfaces contextual health indicators (on track, needs attention, at risk) and explains why. Attempts count in aggregate across the full serve. Risk surfaces early, not after failure.

Unified history

Reassignments no longer reset the conversation. A single timeline preserves all attempts, chats, and documents regardless of how many times the serve changes hands.

Address intelligence

Many failed attempts came from address quality issues. We integrated Melissa Data to validate addresses proactively and surface risks before dispatch.

Client visibility required ops to see the same unified context. Stakeholders proposed redesigning the entire ops platform.

I recommended a narrower approach: a global job selector that surfaces key context without changing existing workflows. Ship value now, defer the full redesign.

“Thank you for listening to our feedback and keeping the workflow very similar while also adding value with information in ways that we can go self-serve it and not confuse the team. Taking our feedback and doing what you did, it's a world of efforts, and it's very much appreciated.”

— VP of Operations

I partnered closely with engineering to define the data model, edge cases, and interaction behavior, reviewing implementations for fidelity and feasibility. I worked with the PM to pressure-test scope, sequence work, and decide what not to build — prioritizing clarity and adoption over feature breadth.

And I collaborated closely with my amazing design team; defining the system-level approach, setting design direction, making scope and sequencing decisions, and ensuring the final experience shipped as a coherent, end-to-end system.

By redesigning the client experience around a continuous story (and enabling internal teams to support that truth), we reduced confusion, restored trust, and created a scalable foundation for retention.