DEV Community

Usman Khan
Usman Khan

Posted on Originally published at usmanakhan.hashnode.dev AI-assisted

The Local Constraints Are the Spec

Most software gets designed in a quiet room with a fast connection, in one language, on a normal working week. Then it ships into a market that has none of those things.

In the Gulf, that gap is where builds go wrong. Arabic reads right to left. Ramadan changes when people work, play and approve things. Data often has to stay inside the country. A tablet in a supermarket aisle may not have a signal. None of this is exotic. It is the environment. Yet teams still treat it as a list of edge cases to patch after launch.

My view, after a lot of delivery across regulated enterprise, marketplaces and retail, is simple. The local constraints are the spec. Design for them on day one, or pay for them twice.

Here is what that looks like in real builds. I will keep the examples anonymised and stick to what the work actually delivered.

Build bilingual in, do not bolt it on

A regulated fractional real estate platform in Saudi Arabia needed a public website that explained fractional investing to both retail and high-net-worth audiences. It had to be fully bilingual, keep all user data inside the Kingdom under PDPL, and hit strict performance and accessibility targets.

The important decision was made early. Bilingual and RTL support were engineered in from the start through locale-aware routing, not retrofitted. Arabic RTL and English LTR were treated as first-class routes, with structured data across both languages. A native Arabic reviewer was on rolling review from day two, not brought in at the end to find what was broken.

The delivered targets were Core Web Vitals of LCP under 2.5s, CLS under 0.1 and FID under 100ms, with WCAG 2.1 AA conformance across both text directions. A QA suite covered 18 end-to-end journeys in both languages.

If you have ever tried to flip a finished LTR layout into Arabic, you know why this matters. Margins, icons, charts, form validation and SEO all assume a direction. Deciding that once, at the routing layer, is far cheaper than discovering it screen by screen.

Treat data residency as architecture

On that same platform, data residency was not a legal footnote. It shaped the infrastructure: KSA-aligned cloud regions, a web application firewall, and encryption at rest and in transit. Keeping all data inside the Kingdom is what cleared a regulatory blocker to launch.

That is the lesson for builders. If a market has residency rules, they decide where your database lives, which managed services you can use, and how your CRM and forms connect. That is an architecture decision. Make it before you pick the stack, not after the compliance review.

Put the calendar in the engine

A multi-sport venue booking platform in the GCC, covering 30+ sports and fully bilingual in English and Arabic with RTL, started from a familiar mess. Bookings ran on phone calls, WhatsApp and spreadsheets. Operators dealt with double-bookings and manual refunds. Players had no way to compare price, distance and availability.

The core was one inventory model that mapped cleanly across every sport, from a full pitch to a swimming lane, with transaction-level slot locking, policy-driven refunds and a ledger-based payments layer. But the part I want builders to notice is this: ladies-only timings, family sessions and Ramadan schedules were built into the availability engine itself.

Not a banner. Not a manual override someone has to remember. The engine.

The platform's results included a 20 to 25% improvement in venue utilisation, 65% fewer double-bookings and scheduling errors, and 60 to 80% less manual admin workload. In my experience, numbers like that come from a model that reflects how people in a market actually book.

The same thinking showed up in a regulatory information system for a multi-country pharmaceutical operation. Its workflows were Ramadan and holiday aware, alongside maker-checker controls where the same user cannot create and approve. If your renewal alerts and approval chains ignore the local calendar, your users will route around them.

Design for the weakest connection

A global health NGO network needed one platform for a global membership, including people in low-bandwidth regions. So low-bandwidth performance and captioned video were designed in, with a dedicated low-bandwidth mode. That mode is what extended usable reach across constrained regions.

Retail floors have the same problem in a different shape. A branded tablet game built for in-store kiosk deployment, a 30-second activation tied to proof of purchase, was given offline functionality for retail environments. A store is not a demo room. If the experience depends on a perfect connection, it fails exactly when a crowd has formed around it.

Test in the real environment, not the flat file

Field marketing teams were approving activation assets from flat artwork and finding the problems only after production: branding that did not read at distance, elements lost against the fixture, formats that did not survive the actual retail environment. Every correction meant a reprint.

The fix was a browser-based mockup tool that drops artwork onto real activation formats, from POSM and gondola ends to kiosks and vehicle wraps, so sign-off moves from flat artwork to in-context preview. It is the same principle as everything above. Check the work against the place it will live, before it is too expensive to change.

A short checklist before you write code

If you are scoping a build for this region, or for any market that is not the one your framework was designed in, ask these early:

  • Language and direction. Is RTL a route from day one, and who reviews the Arabic, and when?
  • Data location. Where must the data live, and what does that rule out?
  • Calendar. Do Ramadan, holidays, family or ladies-only sessions change availability, approvals or alerts?
  • Connection. What happens on a weak signal, or no signal at all?
  • Environment. Where will this actually be seen or used, and have you tested it there?

None of these are features you add later. They are the conditions the software has to survive. AI features follow the same rule: they only help if the system underneath already fits the people and the place using it.

This is the lens I bring to the systems I build, from regulated platforms to marketplaces and field tools. I write more about brand, growth and building at usmanakhan.com.

Build for the room the software will live in, not the room it was designed in. That is usually the difference between a launch and a rework.

Top comments (0)