04 / 05 UAE retail bank Platform-wide design ownership 2023, present

Onboarding as a Product

A customer onboarding platform for a retail bank.

Platforms drift. It is their natural state. Every regulatory update, every product line, every quarterly request pulls the design in a slightly different direction. Hold the line for a year and you have shipped fifty improvements that still feel like one platform. Let go and you have shipped fifty different products dressed in the same wrapper. This is the story of a lending platform that did not drift.

The brief

I owned platform-wide design for the bank's lending stack, the surface customers used to apply, the surface agents used to assist, and the surface ops used to underwrite. The brief was open: keep shipping. The unspoken constraint: keep it feeling like one product.

The problem, defined

The backlog was the problem. Fifty-plus requests, regulatory updates, and bug fixes were live or queued at any time. Each one was reasonable in isolation. Collectively they would have produced a platform with conflicting patterns. Drift would have shown up first in agent error rates, then in customer trust.

The tradeoff that shaped the work

Speed of delivery against structural integrity. The easy answer was to ship each request as a standalone change. The harder answer, and the right one, was to treat the design system as the platform's load-bearing layer and route every change through it. Some features shipped a sprint later. Every feature stayed coherent.

Mapping the experience before mapping the work

Before any backlog grooming, I rebuilt the experience map. One artefact that held every actor, every step, every system touchpoint, and the moments where regulation forces a decision. The map became the contract the squad pointed at when a request looked like it would distort the journey. If a change did not have a clear home on the map, it did not have a clear home in the platform either.

Fig 01 Experience map, actors, steps, system touchpoints, and regulatory decision moments on a single canvas.

What I designed

A lending-shaped extension of the design system. New components, patterns, tokens, and content rules added as each scenario appeared. The system grew, it did not multiply.
Modular flows tolerant of regulatory change. Journeys broken into composable steps so a new disclosure or eligibility rule slotted in without redesigning the whole path.
Mirrored surfaces for customer and agent. Same logic, different density. Self-service was minimal and reassuring. Agent-assist was dense, guided, and built for speed.
A design, QA, dev triage loop. UX defects resolved inside the sprint, not in a follow-up release that would never come.

Outcomes

Fifty-plus improvements shipped without measurable fragmentation.
Agent processing time per application improved by an estimated 15 to 20%.
Customer self-service completion rates lifted on every redesigned flow.
Agent training time reduced through consistent patterns across the platform.

Caveat: directional figures based on internal benchmarks; exact numbers under NDA.

Modularity is the senior craft

At platform scale, modularity is the senior craft. Innovation is cheaper than consistency. Most of the work was holding the line on patterns under pressure, not designing new ones.

Final UI fragments, privilege club, agent workspace, customer ID confirmation, UAE PASS login, and bundled credit card offer. — Fig 02 Final UI fragments across customer and agent surfaces, one component system, two densities.