UX/UI Design
Enterprise SaaS
AI Assisted Workflows
Overview
1
Transaction Screening is a critical part of anti-money laundering operations. Analysts investigate payment screening hits to determine whether transactions should be released, blocked or escalated.
The existing experience had evolved incrementally over many years. Information was fragmented across multiple views, forcing analysts to navigate between screens while manually piecing together evidence before making a decision.
I led the end-to-end redesign of the investigation experience, creating a scalable workflow that simplified analyst decision-making, established reusable interaction patterns and became a key capability showcased during enterprise sales.
Outcomes
Contributed to winning two enterprise customers through demonstrations of the redesigned Transaction Screening experience.
Redesigned the end-to-end analyst investigation workflow, improving usability, consistency and decision-making.
Established reusable interaction patterns that now underpin future AI-assisted investigation workflows.
I led the end-to-end redesign of the Transaction Screening experience, partnering with Product, Engineering, Data Science and AML subject matter experts to translate complex screening workflows into an intuitive investigation experience from discovery through developer handover.
The Challenge
2
01
Fragmented Investigation
Analysts needed to navigate multiple disconnected views to gather enough context before making a decision, increasing cognitive load and slowing investigations.
02
One Workflow, Two Very Different Problems
Transaction Screening had been created by splitting the existing Transaction Monitoring experience.
While this accelerated development, it inherited workflows, terminology and interaction patterns designed for behavioural monitoring rather than live payment screening.
As a result, analysts were forced through a workflow that didn't match how Transaction Screening investigations actually worked.
03
Repetitive False Positives
Analysts frequently investigated the same legitimate transactions because there was no seamless way to convert investigation outcomes into reusable screening rules.
This increased manual effort and reduced operational efficiency over time.
Research Insights
3
Analysts spent more time finding information than evaluating risk
Information required for a single investigation was spread across multiple screens. consumption process
Existing workflows reflected the underlying system architecture rather than analyst mental models
Users were forced to understand how the platform was built instead of how investigations naturally progressed.
Analysts needed confidence before speed
Speed mattered, but only when users understood why a recommendation was being made.
Key Product Decisions
4
Instead of distributing information across multiple tabs, I consolidated investigation into a single workspace where analysts could review hits, supporting evidence and make decisions without losing context.
02 Progressive Information Hierarchy
Not all information is equally valuable during an investigation.
Rather than displaying every attribute with equal emphasis, I organised information around analyst priorities. Frequently referenced evidence remained immediately visible while secondary information was progressively disclosed as investigations deepened.
This reduced visual noise while ensuring analysts could still access supporting evidence when required.
03 Investigation-Driven Allow Listing
Analysts repeatedly investigated the same legitimate transactions because creating allow list rules was a separate administrative workflow. Breaking away from the investigation interrupted the analyst's flow and discouraged proactive maintenance.
I integrated allow listing directly into the investigation experience, enabling analysts to create exclusion rules from the current screening hit without leaving the task.
Rather than treating allow listing as a standalone administration feature, it became a natural extension of the investigation process. Analysts could immediately convert a verified false positive into a reusable rule, reducing repeat investigations while maintaining full visibility over the rule being created.
Reflection
5
The most difficult part of this project wasn't redesigning the interface—it was reducing cognitive load without reducing analyst confidence.
Every design decision balanced speed with trust. Analysts needed enough information to make confident decisions, but not so much that investigations became overwhelming. This project reinforced the importance of designing around user intent rather than underlying system architecture and established patterns that continue to influence the wider platform.
