Designing Clarity in an Anti-Money Laundering Platform

A compliance product used by assessors and admins to evaluate customer risk. The system must meet strict regulatory standards while remaining usable for non-expert users.

My Role: UX Lead
Led research, testing, design direction, and stakeholder alignment.

The problem

The product supported complex AML requirements but relied on:

Long, generic guidance
Hidden risk logic
No structured workflow
No in-product quality control

Assessors were expected to act as compliance experts, despite this not being their primary role.

Who were we designing for?

The Assessor

Assessors complete risk assessments as part of their day-to-day work, but compliance is not their core expertise. They need to review data, interpret risk, and justify decisions.

Without clear structure, they rely on guesswork and external tools. Their goal is to complete assessments efficiently and feel confident their decisions are correct and compliant.

The Admin and AML expert

Admins and AML experts define how risk should be assessed and ensure quality across all cases. They review assessments, guide decisions, and handle complex or high-risk scenarios.

They expect the system to enforce standards and support consistency, not rely on assessors to interpret compliance on their own.

Discovery, Research and Testing (Round 1)

This was a complex problem space with multiple use cases and competing user needs, requiring a flexible and iterative approach.

Interviews + observational sessions

Watched assessors complete real risk assessments
Identified where users got stuck and relied on external tools
Observed QC happening outside the product

Workshops + hypothesis setting

Aligned with stakeholders on key assumptions
Defined core hypothesis:
- Lack of structure and visibility is the main blocker
- Not lack of data or tooling

Low to mid fidelity wireframes

Explored:
- Risk factor visibility
- Assessment structure
- Guidance patterns

Prototype testing – Round 1

Tested end-to-end assessment flow
Introduced AI-assisted elements

What we learnt

AI worked… but didn’t solve the problem

AI outputs were seen as useful
But users still:
- Didn’t know what to do
- Got stuck in the flow
- Lacked confidence in decisions

Core issue identified

The problem was not generating content.
It was supporting decision-making and workflow completion.

Key opportunities uncovered

Admins were already doing quality control externally

Guide assessor decisions with admin configured risk factors and guidance

Ensure compliance quality. Opportunity to bring quality controls into the product and close the loop

Iteration

After running the discovery and testing phase we pivoted to better align with what we were hearing users say.

Shifted from: Improve assessment experience

A general improvement

To: Create a complete compliance workflow

Introduced QC feedback loop
- Admin review
- Request changes
- Provide feedback
Designed for:
- High-risk cases
- Complex scenarios
- Real collaboration between roles

AI repositioned

Not the primary solution
Used to:
- Support admins in configuring risk rules
- Speed up setup, not replace judgement

Validation (Round 2)

Not a full redesign but a shift in focus, adding the quality control feature and enabling admins to support their users within our system.

Key Feedback

Round 2 testing showed clear improvements in structure, clarity, and workflow alignment.

Users were better able to understand risk factors and navigate the assessment process, with less hesitation and fewer points of confusion. The introduction of a quality control flow also aligned closely with how teams already work in practice.

However, gaps remained around decision confidence and clarity of system behaviour, particularly in more complex or high-risk scenarios.

These insights reinforced the need to further support decision-making, improve visibility of system logic, and refine how feedback and actions are presented. They also validated the direction of embedding collaboration and QC directly into the product as a core part of the experience.

Assessor

“I understand what I’m meant to do now, it’s much clearer step by step.”

Reviewer / QC

“This feels much closer to how we actually review cases.”

Manager

“It’s good to have this feedback loop in the system instead of doing it externally.”

Compliance / AML

“I still want to understand why something is considered high risk.”

The Final Design

Not a full redesign but a shift in focus, adding the quality control feature and enabling admins to support their users within our system.

Structured risk assessment flow

Clear progression
Reduced ambiguity
Chunked information
Data organised to match assessor mental model

Visible risk factors

Exposed logic
Clear relationships and impact

Quality control workflow

In-product review and feedback
Closed the compliance loop

Decision support

Admins create guidance for key moments
Improved reasoning quality

What I Learned

Working on this project reinforced that the core challenge in compliance tools is not complexity, but how that complexity is presented and supported.

I learned that users don’t struggle with data, they struggle with making decisions. Without clear structure, visible logic, and feedback loops, even well-informed users lose confidence and consistency breaks down.

This project also highlighted the importance of challenging assumptions. Initial focus on AI did not address the real problem. By grounding decisions in user behaviour, I was able to shift the direction towards workflow, clarity, and collaboration.

Ultimately, the most impactful change was moving from a system that expected expertise to one that actively supports users in making compliant, defensible decisions.

Lesson One

Solving the wrong layer (AI) won’t fix structural problems

Lesson Two

Workflow gaps matter more than feature gaps

Lesson Three

Compliance tools must support collaboration, not just individuals

Lesson Four

Observing real behaviour is more valuable than stakeholder assumptions