Case 01 · pawabloX · Under NDA

One system, seventeen markets

One design system across seventeen markets, shared by design, engineering, QA, research and product — and open enough that anyone in the company could build with it.

Role
Head of Product Design & UX · system owner
When
Apr 2025 — present
Where
betPawa · 17 African markets

Challenge

Seventeen markets and no shared system. The same component existed several times over, and every team described it differently.

Approach

Foundations first, then components, then patterns — one library living in both design and code, with a standard each layer had to meet.

Result

Five disciplines working in one language. Dark theme, once quoted in months, shipped as a single release.

17markets on one system
5disciplines sharing itdesign, engineering, front-end, QA, research
1release for dark themeonce a months-long retrofit

The problem

What was actually wrong

betPawa runs across seventeen markets, and years of legacy had left the same control living in several places at once, behaving differently in each.

The cost was not dramatic, it was constant. Every feature re-argued decisions a system should have settled, and any change that touched the whole surface was quoted in months — so nobody proposed one.

Before

The same control, four times, with three behaviours.

After

VARIABLESCOMPONENTSPATTERNSPAGES
One spine, four layers, five disciplines naming the same thing.

My role

Who did what

Mine

  • Set the direction: systems before screens, foundations before components.
  • Defined the layers — variables, components, patterns, key pages — and the standard each had to meet.
  • Opened the system up in Claude, so people could build with it instead of asking for it.

The team’s

  • The design team built the layers out and owned their product areas within them.
  • Front-end and engineering owned the system on the code side.
  • QA and research defined the states and the evidence it had to cover.

Process

How it went

  1. 01

    Start from what is in production

    We audited the live markets rather than the library. The gap between the two was the real backlog.

  2. 02

    Foundations before components

    Colour, type, spacing and states first. Nothing is visible for weeks, which is exactly why most systems skip this and stay fragile.

  3. 03

    One library, two homes

    The same components in design and in code, so nobody had to translate between them.

  4. 04

    Let everyone build with it

    With the system inside Claude, an idea arrives as something you can click through instead of a document to be explained.

Before

ANYONE WITH AN IDEAWRITES A DESCRIPTIONA DESIGNER INTERPRETS ITA PROTOTYPE EXISTS
Two handoffs, and meaning leaks at both.

After

ANYONE WITH AN IDEABUILDS ON REAL COMPONENTSA PROTOTYPE EXISTS
The design system is the thing they build with.

Decisions

What was chosen, and what it cost

Foundations before a visible component library

Over
Having something to show in the first month
Because
A library on unstable foundations locks the inconsistency in, and every fix after that means moving everything again.
It cost
Weeks with nothing to demo. That is a plan you have to defend twice.

A system people could build with

Over
A design-only library with the usual handoff
Because
It changes who is able to make something. Ideas stopped being described and started being built.
It cost
A second surface to keep in sync, deliberately and on a schedule.

Outcome

What came of it

Five disciplines now use the same names for the same things, so far less gets re-explained at every handover.

Dark theme is the proof, not the goal: months of retrofit became one focused release, four months after the system existed.

Product, UX and tech share ideas as working prototypes now, so customer feedback arrives before engineering is committed.

Lesson

What it cost me to learn

Watching how the team used AI over several months, I saw us looping — prompt, fix, prompt again — and calling it progress. I wrote it up publicly as “The Dark Side of the AI Loop”. The advantage goes to teams who use AI to learn faster, not to produce more.