The Journey

The Journey

The Journey

Two phases. Two different problems. One continuous story.

Two phases. Two different problems. One continuous story.

Phase 01 reduced cognitive load. Officers were relying on memory and personal notes to know which documents a case required, spending 60 minutes on setup before verification even began. We redesigned the system to carry that burden so officers didn't have to. 
Phase 02 introduced intelligence. With the UX foundation in place, we layered AI to automate 80%+ of manual verification checkpoints — taking case time from 3 hours to 30 minutes.

Phase 01 reduced cognitive load. Officers were relying on memory and personal notes to know which documents a case required, spending 60 minutes on setup before verification even began. We redesigned the system to carry that burden so officers didn't have to. 
Phase 02 introduced intelligence. With the UX foundation in place, we layered AI to automate 80%+ of manual verification checkpoints — taking case time from 3 hours to 30 minutes.

The thinking behind each phase

The thinking behind each phase

The thinking behind each phase

Phase 01 - UX

The platform wasn't being used. Officers were working around it.

The platform wasn't being used. Officers were working around it.

12 officers interviewed. All described the same broken workflow — up to an hour of setup before verification could even begin. Case time varied from 2 hours for smaller cases to 4 hours for complex ones — but both shared the same 60-minute bottleneck. Eliminating it had the biggest impact on smaller cases, effectively cutting their time in half.

Improved Experience of User

Fixes Usability Issues

60 min setup → near zero

60 min setup → near zero

Decision Point

UX alone wasn't enough to scale. The system still lacked intelligence.

UX alone wasn't enough to scale. The system still lacked intelligence.

After Phase 01, case time dropped from 4 to 3 hours. Good — but verification still relied on human memory and judgment. Outcomes varied officer to officer. To scale consistently, we needed AI.

Decision: introduce AI

Phase 02 — AI

120+ checkpoints mapped. I validated across 8 trials on LLM. Then we designed.

120+ checkpoints mapped. I validated across 8 trials on LLM. Then we designed.

We didn't design first and hope AI would work. We ran 8 LLM trial rounds, reached ~90% accuracy, mapped every checkpoint to AI vs. human ownership — then built the experience around that evidence.

~90% AI accuracy

120+ checkpoints mapped

Pilot → full rollout

Pilot → full rollout

Result

4 hours → 30 minutes, 2 cases a day to →5 - 6 cases

4 hours → 30 minutes, 2 cases a day to →5 - 6 cases

The combined result of two phases, each solving a different problem. Phase 01 removed friction. Phase 02 removed manual work. Together: ↓88% time reduction, 3× more cases per officer per day.

↓88% total time reduction

3× cases per day

Document Verification - Phase 01 & Phase 02 (Case Study

Document Verification - Phase 01 & Phase 02 (Case Study

Document Verification - Phase 01 & Phase 02 (Case Study

Senior UX Designer · Central Ops · 2024 - 2025

Senior UX Designer · Central Ops · 2024 - 2025

Senior UX Designer · Central Ops · 2024 - 2025