Leading South African Investment Bank

Building Design Infrastructure at Scale

A case study in systems thinking at enterprise scale.

Design systems don't fail because of bad components. They fail because nobody owns the hard part, the alignment, the governance, the organisational change.

Over more than a decade at one of South Africa's most respected investment banks, I built, lost, and rebuilt design infrastructure, each time at greater scale. The design systems work ran as a continuous thread across every role, until it became a dedicated function in its own right.

The bank's digital presence had grown organically: inconsistent experiences, disconnected visual languages, no shared standards. Leading the Marketing design practice, I built a complete design system alongside broader team and brand responsibilities — without a dedicated team or mandate.

Eventually a wireframe alone gave developers everything they needed to build a page.

• Atomic foundations: typography, colour tokens, spacing, iconography and grid systems
• 20+ core components built in Adobe XD, migrated to Figma
• Flows and templates for every major page type
• A design-to-development bridge that cut delivery from weeks to days

“Shane has done by himself what our competitors do with teams of 20+ resources. He designed a digital brand and his toolkit will truly change the way we think about digital within the bank.”

Marketing Technology Platform Lead

As roles shifted, ownership of the system became unclear. Without a dedicated steward it degraded — components unmaintained, documentation outdated, consistency eroded.

This was not a failure of effort. It was a structural problem: infrastructure without formal ownership is always at risk.

A design system is not a product you ship, it is living infrastructure. This understanding shaped everything that came next.

The role eventually became formal: Head of Design Systems and Design Practice Lead. The mandate expanded from one practice to the entire organisation — multiple brands, multiple front-end stacks and the challenge of making the case for infrastructure investment inside a project-oriented culture.

Many platforms across the bank were built independently, each with its own frontend and no shared standards. They needed a system now, not a longer-horizon solution they couldn't yet build on. The work had to serve them immediately, practically, without constraint.

At group level, a shared frontend standard is being developed that will shape every digital product the bank builds for the next decade. That effort is weighted toward retail banking. Institutional banking is structurally more complex, and without active representation, that complexity will not be reflected in what gets built.

The strategy is to build a foundational system and use it as the vehicle to influence the group direction. The group-level team has agreed to work together toward convergence: ideally a single unified system, or at minimum a shared token foundation underneath both.

AI tools are already writing code. The question is not whether teams will use them, it is whether the foundation those tools build on is coherent enough to produce something worth keeping. AI has infinite execution speed but zero inherent judgement. The design system provides that judgement, constraining output to what is correct, consistent, and on-brand.

Components are being built with structured documentation and defined contracts that AI systems can read and act on, not just humans. Naming conventions, token architecture, and component APIs are all designed for machine readability as well as developer use. The governance layer is not a constraint on AI adoption. It is what makes AI generation predictable, consistent, and worth building on.

• Advanced token architecture: light and dark modes, density, prominence, responsive scaling, inverted schemes, per-brand colour theming
• 25 foundational components in Figma, moving into development as web components
• Multi-brand capability demonstrated across multiple brands
• Component documentation structured for both human and AI consumption
• Active adoption requests confirmed across multiple business units, before the system is complete
• Collaborative engagement established with the group-level team toward a converged token foundation

Organic adoption before launch is the strongest early validation a system can get.

Real, adopted design infrastructure, delivered alongside broader responsibilities, at enterprise scale.

Understanding why systems fail structurally, not just technically and using that to build better ones.

Making the organisational case for infrastructure investment and building alliances, without a team or a mandate, until one was earned.

Designing for how humans and AI agents will build together, with the governance layer ready when they do.

This work sits at the intersection of design, technology and organisational change. The technical build is the easier part. The harder work is making the case, earning the mandate and keeping the system alive once it exists.

Design systems work when someone owns the hard part. The alignment. The advocacy. The long game.
That is the work I do.

“You don't rise to the level of your goals. You fall to the level of your systems.”

James Clear