CSC was undergoing a full digital transformation, redesigning every member-facing product end to end. The organisation had decided to build a UX function for the first time. I was brought in to lead that establishment while delivering product outcomes across multiple squads simultaneously.

That meant operating at two altitudes: shipping product improvements with hard outcome targets, while building the infrastructure, standards, and organisational credibility that would make UX a strategic function rather than an execution service.

CSC's retirement planning tools were built for financial experts in a domain where expertise is rare. Most members only engage seriously with their super at the point of retirement transition, often for the first time, with no framework for the decisions they needed to make.

Qualitative interviews with members aged 45–65 across PSS and MSBS schemes, combined with quantitative behavioural analysis of existing tool usage. Four behavioural patterns driving poor outcomes.

Gallery — research artefacts

Calculator errors reduced 60%. Task completion improved 45%. RIS modeller scored 4.1/5. Design system adopted across 12 squads with 30% improvement in development velocity.

Gallery — shipped product

The strongest decision was the earliest: refusing to accept that this was a usability problem. The tools were not hard to use because of poor interaction design. They were built for the wrong user. What I would do differently: include members who had already transitioned into retirement in the research cohort. That group would have validated whether the modeller's outputs mapped to real-world outcomes, not just usability satisfaction scores.

Flinders was undergoing a full web transformation. As sole UX lead within the CX team, I ran a multi-stream research programme covering two critical stages of the student journey: course discovery and enrolment onboarding.

Both stages shared the same root cause: the digital experience was built around institutional logic, not student mental models.

Four parallel research streams ran across both stages of the journey.

Gallery — research artefacts

International prototype outperformed current pages by 35% on task completion and 39% on discovery. Domestic course recommendations in phased rollout. Onboarding research produced five named findings and a four-phase product roadmap accepted by cross-functional teams.

Gallery — shipped work

The contact centre data was the most underused asset in the organisation before this project. Integrating it as a quantitative layer changed how stakeholders received the findings. What I would do differently: establish a continuous feedback loop between contact centre data and the product team from the start of the transformation, not mid-stream.

Teams across marketing, product, and content were waiting weeks for participant recruitment before they could test ideas. Teams facing launch deadlines shipped without testing. Assumptions went unvalidated. When research did happen, it was often too late to change the direction it surfaced.

Conducted 60+ real participant interviews across six distinct student cohorts before building anything. Each cohort mapped across pathway logic, content needs, support requirements, and decision triggers. This data became the foundation the synthetic personas were built from. Not a set of assumptions, but a structured distillation of real interview evidence.

Gallery — interview and research process

Adopted as the institutional research framework across the CX team. Teams now pressure-test concepts without waiting for participant recruitment, removing a 2–4 week bottleneck from the early-stage research cycle.

The responsible usage guidelines are in active use, treating synthetic personas as hypothesis generators that require real-user validation before high-stakes decisions ship.

The most important thing I built was not the platform. It was the responsible usage framework. AI-generated personas are easy to over-trust, especially in organisations where research capacity is thin. What I would do next: build a validation loop into the platform itself, so teams are prompted to schedule real-user testing when a concept moves past the hypothesis stage.