UX CASE STUDY · GOVTECH · 2021 - 2024

Cutting court audio handling time 30% for staff who get pulled away mid-task.

A desktop hub that let Supreme Court staff process audio requests without losing their place when a courtroom called them away. It cut overall handling time about 30% and won the 2024 Client Services Excellence Award.

SPEC · TWM001
Role
UX Designer & Developer
Timeline
2021 - 2024
Stack
Python · BeautifulSoup4 · Playwright · Flutter · Figma
Constraint
100% LOCAL, OPEN SOURCE
HANDLING TIME (INDEX)AFTER vs BEFORE 100→70
MY ROLE002

I was a court clerk on the In-Court Technology team, not a hired designer. I built this because I lived the problem daily and nobody else was going to fix it.

Team
Solo build, In-Court Tech team
Stakeholders
Court clerks, managers, IT
FIG. 012021 - 2024
PROBLEM → SOLUTION003

The Problem vs. The Solution

Live court always took priority. Staff processing audio requests at their desks were constantly pulled into courtrooms, and the multi-tool process meant every interruption threw the work away.

BEFORE003.A

Requests died on interruption

The existing process for retrieving audio, editing log notes, and filtering files was slow, manual, and spread across multiple tools.

P-01
Courtrooms interrupt everythingLive sessions always outranked desk work. A clerk mid-request would leave for a courtroom and the request would sit abandoned.
P-02
Log notes edited by handCourt log notes lived in HTML tables. Isolating one file's rows meant several minutes of manual editing in tools like Paint.
P-03
One request, many toolsRetrieving audio, filtering files, and preparing exports each lived in a different tool with its own steps.
AFTER003.B

One hub that survives interruptions

A desktop app that kept requests processing while staff were away, and collapsed the multi-tool workflow into one place.

S-01
One-second log-note filteringBeautifulSoup4 finds the file numbers inside the HTML tables and rebuilds a filtered log note automatically. Minutes of editing became about one second.
S-02
Single-click request pipelineType a file number and the app pulls the transcript order, filters the files, and exports everything in one click.
S-03
Built to keep workingRequests kept processing while staff were called away. The final version was rebuilt in Python and Flutter with layouts designed in Figma.
THE IMPACT004

The Impact

Measured against how the work actually happened before, across three shipped versions.

Less handling time
0%
Across the full request workflow
Log-note filtering
~0s
Down from several minutes of manual editing
Local, open source
0%
No cloud, no third-party data
RETURNED TO STAFF, EVERY REQUEST004.B
0MIN
A 15+ minute manual process down to about 2, before counting the restarts that interruptions used to cause.

Request handling time

Indexed. Before the tool vs. the final Flutter version.
KEY DECISIONS005

Key Design Decisions

Three calls that shaped the tool, each made against a real constraint rather than a preference.

DEC-01
The 50% rule

No feature shipped unless it cut the task by roughly half the clicks.

Why

Anything under 10% would not justify asking staff to stop everything and relearn a system they already knew. The bar kept the tool honest.

DEC-02
Local and open source only

Everything ran on local machines with open-source libraries. No cloud, no third-party data sharing.

Why

Court data is sensitive and policy review kills projects. Staying local removed the security question entirely.

DEC-03
Remove the Playwright feature

V2 pulled transcript orders from the court's internal web service. I removed it.

Why

The internal service was being updated and the integration would have broken in production. Reliability beat features.

OLD vs NEW WORKFLOW006

The Same Request, Both Ways

Every step a clerk walked through under the old process, against the hub. The old side is also where a courtroom interruption meant starting over.

OLD WORKFLOW · 15+ MIN PER REQUEST006.A
  1. 01Open the request and note the file number
  2. 02Locate the hearing audio in the recording system
  3. 03Export the proprietary audio files
  4. 04Open the court log notes as an HTML table
  5. 05Manually isolate the file's rows in an editor like Paint
  6. 06Rebuild a clean log note from the fragments
  7. 07Prepare and stage the files for export
  8. 08Assemble the request package by hand
  9. 09If a courtroom calls, abandon everything and restart later
NEW WORKFLOW · ~2 MIN PER REQUEST006.B
  1. 01Type the file number into the hub
  2. 02Review the auto-filtered log note and staged audio
  3. 03Export the complete package in one click
9 STEPS → 3 STEPS · −67% · 15+ MIN PER REQUEST → ~2 MIN PER REQUEST
DESIGN EVOLUTION007

The Agile Evolution

Twelve months of iteration condensed into one viewport: from MS Paint sketches to a shipped Flutter application, across three versions tested with the clerks who used them.

Understanding the user and building the MVP.

Due to strict security, the first iteration had to be built using tools already available. I gathered feedback from transcript clerks through user interviews and journey mapping, then sketched the interface in MS Paint to map out the logic for a functional Python/Tkinter application.

  • ▸Conducted user interviews and journey mapping
  • ▸Heuristic evaluation of current tools
  • ▸Created low-fidelity wireframes
BEFORE / AFTER008

Three years between these two screens

Drag the handle. The Tkinter MVP that proved the idea against the Flutter hub that won the award.

V3.0 · Flutter Hub
V1.0 · Tkinter MVP
REFLECTION010
The tool people opened without being told to

The best signal was not the award. It was walking past desks in the morning and seeing the app already open, unprompted. Later, at BrainStation, I found out the things I had been doing by instinct, shadowing users, setting adoption bars, testing with real clerks, were already the UX methodology. This project is why I trust adoption over praise as the measure of whether a design works.

NEXT CASE STUDY

Secure AI Transcription

▸
Available for Work

Let's build something impactful.

I'm currently open to new opportunities in UX/UI Design, Product Design, and GovTech transformation roles.

Drop me an email Connect on LinkedIn