Most UAE businesses say they use AI. Few pay for it.
Four independent studies show the same gap between saying you use AI and getting something from it. This is what I think sits in that gap, and why I work the way I do.
By Marcus Cent5 min read
Two figures about the same country, published in the same year, do not sit comfortably together.
A survey by AWS and the UAE Artificial Intelligence Office found that 72% of businesses say they have adopted AI. Pemo, a UAE spend-management company, looked at card and bill payment data from more than 6,000 UAE businesses, January to March 2026, and found that 12% are paying for an AI tool, measured by actual transactions.
The two are not measuring the same thing, and I would not subtract one from the other. The first figure is a survey. The second is payment data from one company's own customers, so it is not a national figure. But the direction is hard to argue with. Saying you use AI is easy. Somebody in the office has a free chatbot open in a browser tab, and that may well count. Paying for a tool is a decision, and far fewer businesses have made one.
The same pattern worldwide
This is not peculiar to the UAE. McKinsey's The state of AI in 2025, which had 1,993 respondents in 105 countries, surveyed June to July 2025, found that 88% of organisations use AI in at least one part of the business. In the same survey, 39% can attribute any effect on profit to it.
BCG reached a similar conclusion in Where's the Value in AI?, based on 1,000 senior executives in 59 countries and published in October 2024: 74% of companies have yet to show tangible value from their use of AI.
Different firms, different methods, and the same shape each time. Adoption is high. Results are much harder to find.
Adoption is easy to claim. Value is hard.
I don't think the gap is about the technology. The tools are good and they keep improving. The gap is knowing where, in one particular business, AI changes a cost, a delay or a lost sale.
A tool cannot answer that for you, and I cannot answer it from a desk. It depends on how the work really moves through your business: who retypes what, where an enquiry waits for a reply, which quote goes out late because the person who knows the price is out on a job. Those things are rarely written down. Often the owner knows roughly where the trouble is but has never had the time to trace it.
A general-purpose tool knows none of this. A chatbot subscription gives each member of staff a capable assistant and leaves the business itself working as it did before. That is how a company can honestly answer "yes, we use AI" in a survey and still see nothing change in its accounts.
I started my first company, Marcat, in Glasgow in 1997, and I have been building software for businesses ever since. I co-founded ResDiary, which was acquired by AccorHotels in 2018. ResDiary did one thing that restaurants do every day, which is take reservations. The lesson I took from it is older than AI: software earns its keep when it is built around a specific job that a specific business does every day. AI has not changed that. It has changed how quickly I can build software of that kind.
Why I start with a visit
So I start by coming to see the business. I would rather stand in your office, workshop or warehouse and watch how a job moves from enquiry to invoice than write a proposal from a phone call.
Two businesses in Dubai, Powerworks Garage and The Garage Guys, run on software I built. I should say that I have a stake in both: I am a founding investor in Powerworks Garage, and I founded The Garage Guys. In each, the useful part came from knowing the business. At Powerworks a car is booked, checked in, inspected, estimated, approved, worked on, invoiced and paid, and each step hands over to the next without anyone retyping anything. AI reads the vehicle registration cards. At The Garage Guys an AI design studio turns a site photo into a design concept, which leads on to the proposal and the contract.
In both, AI does a specific job inside a system that fits the business. I could not have chosen those jobs without seeing how the work was done.
Why I build before I ask for money
The second habit follows from the first. If most companies cannot yet show value from AI, an owner is right to be sceptical of anyone selling it, and that includes me. A slide deck does not settle the question. Working software does.
So when I can see a clear problem worth solving, I build a working first version before I ask for any money. I call it the Proof Build. It is one working piece of software that solves one real problem in your business, built on sample data and shown to you at your premises. If it works and you want it in production, you get a fixed written quote. If it does not, you have paid nothing, and we both know more about where the idea falls short.
I cannot tell you from here whether AI will pay for itself in your business, and neither can a survey. What I can do is come and look.
Sources
- McKinsey, The state of AI in 2025. 1,993 respondents in 105 countries, surveyed June to July 2025.
- AWS and the UAE Artificial Intelligence Office, 2026: 72% AI adoption rate across UAE businesses. Pemo, 2026: UAE businesses and AI spend. The first figure is a survey. The second is card and bill payment data from more than 6,000 UAE businesses, January to March 2026.
- BCG, Where's the Value in AI? 1,000 senior executives in 59 countries. Published October 2024.