In October 2017 the United Arab Emirates appointed the world's first Minister of State for Artificial Intelligence — years before most global boardrooms could say what the phrase meant. That head start matters to anyone building in the Gulf right now, because the region did not stumble into the AI moment. It planned for it, and the institutions that will buy your software have been preparing for a decade.
Which creates an odd mismatch. The tools have made prototypes almost free, and the buyers have become more demanding at exactly the same time. Regular readers will know I have never been persuaded that speed of building is the interesting variable; what a thing is worth six months later has always been the harder question, and it is now the only question that separates two products that look identical in a demo.
The prototype that cannot be sold
I meet the same founder several times a year, in Dubai and in Riyadh, and the conversation runs almost word for word the same way.
They have a product. It works, customers use it, revenue is real. Then an acquirer or an investor asks for a technical review, and within a fortnight the deal has either repriced or gone quiet. Nothing was broken. The software still does what it did on the day the term sheet was signed.
What failed was the transfer test. A prototype is something that proves an idea; an asset is something another party can own, operate, insure and sell. The gap between those two is not code quality — plenty of generated code is perfectly decent — it is everything around the code that nobody generated.
And that gap has a price tag, because a buyer who cannot verify what they are buying does not walk away. They discount.
Why "raise first, then build properly" is the wrong order here
The standard advice given to Gulf founders is to prove traction, raise capital, then use the capital to hire a team who will professionalise the product. It is sensible advice and it works in markets where the buyer is a consumer.
It fails in this region for a structural reason. The most valuable early customers in the GCC are banks, government entities, telcos and large family conglomerates — and every one of them runs a vendor assessment before signature. So the professionalisation you were planning to fund with the raise is the thing standing between you and the revenue that justifies the raise.
What nobody tells you is how small the fix usually is at that stage. Not a rewrite. A fortnight of unglamorous work, done before the first enterprise conversation rather than during it.
The second piece of standard advice that fails is "get to product-market fit, worry about architecture later." I understand why people say it, and for a consumer app in a large single market it holds up. Across the GCC it does not, because fit is not one market — a product proven in the UAE meets Saudi residency requirements, Kuwaiti payment rails and Arabic-first expectations the moment it tries to grow, and a codebase with those assumptions hard-coded has to be opened up at precisely the moment you are least able to pause.
What turns an AI prototype into a software asset?
A software asset is a system whose value survives the departure of the person who built it. Four properties make that true: transferable ownership, documented architecture, operable infrastructure, and evidenced compliance. A prototype has none of them by default, and each can be added in days rather than months if it is done before the codebase grows.
Look at the effort column. The whole list is under a month of one competent engineer's time, and it routinely moves a valuation by a multiple. I know of no other work in software with that ratio.
| Check who legally owns the code before you check anything else. If a freelancer, an agency or a co-founder who has since left contributed without a written assignment, your company may not own its own product — and under the copyright regime recorded by the World Intellectual Property Organization, that transfer has to be in writing. Fixing it after a term sheet costs far more than fixing it now. |
How do investors and acquirers value AI-built software in the Gulf?
Gulf investors and acquirers value AI-built software on transferability, not on how it was written. The questions asked in diligence are who owns the intellectual property, how long a new team needs to take over, whether the system can be operated to a service level, and whether data handling satisfies UAE and Saudi requirements. A codebase that answers all four is priced as an asset; one that answers none is priced as a customer list.
That last sentence is the part founders find hardest to hear, and I would defend it in a room. A buyer who cannot verify a system will pay you for your revenue and rebuild your product, which is a perfectly rational thing for them to do and a bad outcome for you.
Here is the version of this I watched most closely. A logistics startup in Dubai had built its dispatch platform with an AI builder over five months and had genuine traction — real fleets, real money, growing weekly. During diligence the acquirer asked one question: how long would it take a new team to make a safe change to the routing logic? The founder was honest and said he did not know, because he had written all of it himself and had never watched anyone else try. So we tested it. We put two engineers who had never seen the codebase in a room, gave them a small, well-defined change, and timed it. Nine days. That number, on its own, moved the conversation more than any document we produced — and after four weeks of documentation and cleanup, the same test ran in under two days, a reduction of roughly 78%.
| A prototype is worth what it proves. An asset is worth what someone else can do with it after you have gone home. |
Where D33 changes the arithmetic for founders
The Dubai Government announced the D33 economic agenda in January 2023 with a ten-year target to double the size of the emirate's economy. Doubling anything over ten years implies roughly 7.2% compound growth a year, which is a pace you reach through projects rather than press releases — and that is why every institution in the city now works backwards from 2033.
Consider who the buyer becomes under that plan. Digital services procured by government entities through programmes coordinated by Digital Dubai, pilots run through the Dubai Future Foundation, financial products regulated inside the Dubai International Financial Centre, and startups scaling out of programmes like Hub71 in Abu Dhabi. None of those buyers are early adopters who forgive a broken audit trail.
The federal picture reinforces it. The UAE's national artificial intelligence strategy, published on the UAE Government portal, runs to 2031 and treats AI as national infrastructure rather than a sector. Infrastructure implies standards. Standards imply evidence.
So the founders who will benefit most from this decade are not the ones building fastest. They are the ones who make their fast-built products legible to an institution — and I would rather be six weeks slower and legible than first and unbuyable.
What to do in the next thirty days
Five steps, in order, and the first two cost nothing but attention.
- Move every account into the company's name. Repositories, cloud, domain, design files, app store listings, the analytics dashboard. If any of it sits under a personal login, that is the first thing diligence finds.
- Get the assignments signed. Every contributor, including friends who helped for a weekend in the early days, and including the agency you used for two months last year.
- Run the transfer test. Give an engineer who has never seen the code a small, real change and time them. Whatever number comes back is your honest score, and it is the number an acquirer will discover eventually.
- Write down the six decisions. Not full documentation — the six architectural choices a new person would otherwise have to reverse-engineer, each in a paragraph, with the reason.
- Answer the residency question in writing. Where the database, the backups and the logs live, and whether that satisfies the biggest customer you intend to win next year.
Most of what a team like ours at NICGULF does sits inside exactly that thirty days, using engineers from IndiaNIC's delivery hubs who have run the transfer test on other people's codebases often enough to know what it will say before it says it.
The mistake I would most like to save someone from is one we made ourselves. On an engagement two years ago we produced beautiful documentation for a client's platform — diagrams, decision records, an onboarding guide — and never tested it on a real newcomer. Six months later a new engineer joined and took eleven days to make his first safe change, because the documents described the system as designed rather than as built. Documentation you have not tested on somebody is a hypothesis, not an asset. We now measure onboarding by timing it, every time, and it is the only documentation metric I trust.
So here is my question, and it is the one I would ask in your own management meeting this week: if your most knowledgeable engineer took a month's leave starting tomorrow, what could the rest of the team not safely change? Write the list down and count it. Then bring it to the table — because that list is the exact distance between your prototype and an asset somebody wants to buy.
Frequently asked questions
What is the Dubai D33 economic agenda?
The Dubai Economic Agenda, known as D33, was announced by the Dubai Government in January 2023. It sets a ten-year plan running to 2033 with the headline goal of doubling the size of Dubai's economy, alongside targets for foreign trade, investment and the growth of the digital economy across the emirate.
Can software built with AI tools pass investor due diligence?
Yes, provided the surrounding work is done. Diligence examines ownership, maintainability, operations and compliance rather than how the code was authored. AI-generated systems routinely pass once intellectual property is assigned in writing, architecture is documented, deployment is automated and data residency is evidenced — usually a few weeks of preparation.
Who owns the intellectual property in AI-generated code?
Ownership follows the contract, not the tool. Under UAE Federal Decree-Law No. 38 of 2021, software is protected as a copyrighted work and economic rights transfer only by written assignment. A company therefore needs signed assignments from every contributor — employees, freelancers and agencies alike — regardless of how much of the code a model produced.
