I helped guide a multidisciplinary team to reimagine how ANZ's brand could live across digital, product, and motion. We started by taking stock of what worked and what didn't across existing brand touchpoints. From there, we developed a unified visual language built around clarity, warmth, and flexibility.

We refined typography, colour, illustration, and motion principles to feel both cohesive and full of character. Collaboration was key, we worked closely with internal teams and agency partners to ensure the new system could scale and inspire across every channel.

The result wasn't just a refreshed look, it was a living identity that gave ANZ Plus its own confident voice within the broader brand. The new system made creativity faster, more consistent, and more expressive, setting a new standard for how the brand shows up across all our channels.

Four systems, 1.2M+ component insertions and dozens of teams, served by a team you could fit around one desk. That ratio only works if the system is the easiest way to ship, so nobody goes looking for a workaround. Every component we put out has to be quicker to use than to rebuild, and trustworthy enough that design, engineering and brand stop second-guessing it.

It pays off: designers who adopt the system spend 87% less time on design tasks, which gives each of them roughly two workdays back a week.

I don't run the system from a roadmap deck. I'm in the Figma libraries most days, because the quickest way to earn a team's trust is a component that already does what they need.

Every platform gets a full set of templates, mirrored so a designer moving from iOS to web finds the same flows in the same places. I shipped the monogram live text component (15 variants, 5 sizes, 3 platforms), swapping rasterised PNG symbols for text that binds to tokens. I'm also leading the move to a three-tier token architecture and a federated model, so contribution scales past the core team instead of queueing behind it.

A small team has to be faster than its headcount. I built an AI-assisted workflow (Claude Code with custom ANZ design skills) and use it on live deliverables. A table component went from a 120-variant matrix to a row-based build, with instance swaps across 81 cells. Assessment and documentation that used to take hours now takes minutes, which leaves the team's time for the decisions only people should make.

Invisible also means nobody waits on the system when the hardware changes. Our iOS system is built on Apple's native components, so when iPhone Duo arrived with its foldable screen, every component already knew how to respond to the new layouts. ANZ shipped support for it on launch day. No rebuild, no catch-up sprint, and most customers never knew there was anything to do.

A system nobody notices is easy to call a cost centre. So we measure it.

The system has become shared infrastructure that silently speeds up delivery. Most teams only notice it when it's missing.

Before touching a model, we looked at how the current style actually performed. In February 2026, 400 participants across four segments rated the home screens of six Australian banks on five attributes. ANZ led on modern, and ranked first of six on "Unique & Stands Out" at 77.3%. Trust and professional appearance were the clear gaps.

That reframed the problem. The playfulness leadership worried about was also our biggest asset. Strip it out and we'd fix one score by sinking another.

Over a few weeks we pushed three directions across the same set of assets (piggy bank, house, wallet, security shield and so on) and put them side by side with the current artwork. Each came with an honest read on what it would cost to roll out.

Presenting trade-offs instead of a single hero concept meant the design authority was choosing between real options, with the costs on the table. They backed option 3.

A style decision is easy. Applying it to a library of roughly 800 illustrations is where these projects stall. A spreadsheet review or a Figma comment thread would have taken weeks and lost the thread halfway.

We started with a simple framework: two axes, retro and playful to modern and restrained, decorative to informative. Then I built a small tool in Claude that pulls every asset straight from Figma through its API and plots it on that grid. Where an asset lands suggests the action (keep, evolve, revise or retire), the table logs live, and it exports to Excel.

The bottleneck wasn't taste, it was throughput and alignment. The scatter plot let the team argue about where something sits on a spectrum instead of me ruling on 800 yes-or-no calls, and the export became the source of truth for what got reworked first, plus the evidence for stakeholders when something needed retiring.

Each change answers a specific part of the brief. The final renders differ more subtly than you'd expect, on purpose: evolve the style, don't reinvent it.

A style decided in a workshop doesn't survive contact with a whole design org unless it's written down. So we built the ANZ Visual Identity Playbook: 48 slides that cover where each kind of imagery belongs, and how the new 3D artwork gets made.

The usage half starts from one rule: nothing is decoration. Photography leads product discovery, icons do the scanning, and 3D illustration is saved for moments that deserve to feel like one, like onboarding milestones or a campaign header. Servicing emails such as security alerts get no illustration by default, because warmth is the wrong signal when you're telling someone to check their account. The production half walks through concept, modelling, texturing and lighting, and compositing, with do's and don'ts at each stage, including a fixed set of around ten textures so material choices stay as disciplined as the colour.

We tested before locking anything in. In May 2026, 402 participants across youth, migrant, affluent and business segments saw old and new versions of five designs and picked a favourite. The new style won four of the five, most clearly on Getting Ahead and Additional.

Burst lost. People preferred the old version 46% to 32%, and it lost in three of the four segments, which reads to me as a problem with that execution rather than the direction. On attributes, the new work scored best on trust (3.71) and modern (3.70), the scores we set out to move. Unique (3.52) and memorable (3.42) came in softer, so distinctiveness is the thing to keep watching as the style spreads.