Evidence
Case studies make claims. This page is the evidence behind them, mapping my work to what hiring managers typically ask.
AI
Designing for and with AI
Dispute Management
The product handles messy, unstructured input and prepares each case for investigation. I kept the logic visible so investigators can see how the system reached its conclusion.
Prototype-first
Turning ideas into working prototypes so I can test real logic, states, and data early.
Concept exploration
Exploring workflows, edge cases, and system behaviour before committing to design direction.
Simulated persona reviews
Stress-testing screens and workflows, challenging assumptions, and surfacing ambiguity or cognitive overload.
Design audits
Reviewing component and workflow consistency across all products in the procurement platform.
Research synthesis
Synthesising complex inputs and articulating trade-offs.
Complexity to clarity
Turning fragmented workflows into a cohesive product
Dispute Management
Investigation time dropped from 45–90 minutes to under 10.
Rate Management
Rate upload time dropped by 65%, and time to extend or expire rates dropped by 70%.
Port-call Management
Terminal onboarding time dropped from 26 days to 5 days.
Systems thinking
Understanding how parts interact within a whole
Procurement platform
Shaped standardised UI patterns across Maersk’s procurement applications, not just within a single product.
Rate Management
Repositioned rates as a governed source of truth that downstream systems consume, rather than data each system manages locally.
Ambiguity
Designing when the right answer isn’t obvious
Dispute Management
Treated incomplete data as the default scenario, not an edge case.
Port-call Management
Pushed the product to explain and resolve decisions, not just generate POs.
Trade-offs
What I cut, and why
Rate Management
Kept Excel for bulk entry in the first release rather than rebuilding it. It was already the team’s validation tool, and replacing it would have delayed the release.
Dispute Management
Aimed for a single source of truth, settled for a coordination layer because some upstream data was outside our control.
How I’ve changed
Positions I’ve revised when reality pushed back
Dispute Management
Started by trying to remove human judgement from the system. Shifted to supporting judgement when the data turned out to be too incomplete to automate.
Rate Management
Argued for in-UI rate maintenance and lost the call for MVP. I agree it was the right MVP decision, but I still think the long-term value hasn’t been realised.
More work
I turned an invoice comparison tool into a decision system
Investigation time dropped from 45–90 minutes to under 10.
I rebuilt rate handling and terminal setup as tracked workflows
Terminal onboarding time dropped from 26 days to 5 days.
