← All work
01 / 05 UAE retail bank Concept lead 2024 / 25

Banking as a shopfront

Most bank websites are built like internal org charts. You can feel the boundaries between divisions as you click through them. Customers don't think that way. They think about life moments, the new car, the house, the holiday, the rainy-day cushion. This concept rebuilt a regional bank's product discovery on that observation, using AI to translate the bank's language into the customer's.

The brief

A regional UAE retail bank wanted to lift conversion on its product website. The ask arrived as a redesign brief. The work became a re-framing: stop treating the site as a brochure for divisions, start treating it as a shopfront for intents. AI was the lever that made the re-framing affordable, because translating compliance copy by hand was the bottleneck that had killed every previous attempt.

The problem, defined

Site analytics showed visitors entering the product index and leaving before they reached a single product page. Click depth was shallow because the taxonomy didn't match the language customers used. The bank's site read like a sitemap of internal divisions. Customers read it as a maze. Internal stakeholders, product owners, compliance, marketing, each had their own preferred entry point, and the navigation showed every one of them.

Benchmarking, not borrowing

I audited fifteen banks across the GCC and the wider region against four lenses: product discoverability, AI integration, content register, and the gap between marketing surface and product detail. Most banks competed on sameness. A handful had begun to wire AI in, but as bolt-on chatbots rather than as part of discovery. The pattern was clear enough to argue from.

A benchmarking wall of fifteen regional bank websites, annotated by quick wins, enhancements, and gaps in AI integration.
Fig 01 Competitive teardown of fifteen regional banks, sorted by quick wins, enhancements, and the gap in AI-driven discovery.

The tradeoff that shaped the design

Compliance-driven content against customer-driven content. Regulators require certain disclosures and a certain register. Retailers know that disclosure language buried inside a product page kills conversion. The concept routed compliance copy through an AI layer that translated obligation into intent, keeping the legal substance but changing the surface. Legal sat in the working sessions; nothing reached the prototype until they had signed off on the translation rules.

What I designed

  • A shopfront, not a directory. A re-categorised product set framed around customer intents, saving, borrowing, protecting, investing, with filters, side-by-side comparisons, and eligibility highlights up front.
  • AI inside discovery, not bolted on. The assistant lives in the navigation, not in a corner chat bubble. Its job is to translate jargon, recommend against declared intent, and explain eligibility in plain English.
  • Personalisation against intent, not demographics. The system reads what the customer says they want, not what a CRM segment predicts. Cleaner privacy posture, better recommendations.
  • A service blueprint as the spine. Content, product data, AI logic, and UX touchpoints linked into a single map. The concept stayed buildable because the blueprint kept everyone honest about hand-offs.

Measuring AI against jobs-to-be-done

The hardest argument inside the bank wasn't "should we use AI", it was "how do we know it's earning its place". I framed evaluation around the jobs the customer was trying to do, not the features the AI could perform. For each JTBD, find the right product, understand if I qualify, compare options, get unstuck, I scripted a user story, walked it end-to-end in the prototype, and recorded the resulting flow as the success benchmark. The team now has a video library of what good looks like before a single line of production code is written.

Fig 02 One JTBD walk-through, AI guidance answering "can I afford this car" from a single declared intent.

Outcomes (projected)

  • 20 to 30% improvement in product discoverability against the existing benchmark.
  • Plain-language replacement of jargon projected to lift completion 10 to 15% on key product flows.
  • A scalable blueprint for AI integration across channels beyond the website, app, branch tablets, contact centre scripts.

Caveat: directional estimates based on comparable engagements. Concept project, not yet in production.

Where the leverage was

Banks usually copy banks. The strategic move was to copy retail. Most banking sites compete with each other for sameness, which means the competitive advantage was in not playing that game.