Dubai has set the target. The private half has to be built inside the companies.

Published 3 October 2026

Since 4 May 2026, agentic AI in Dubai has been a two-year programme for the private sector, not a trend to watch. This page sets out what the agenda commits to, what is already being delivered, where the remaining half sits, and how we think it gets built inside an owner-run company. Every figure is linked to its public source.

The commitment

The Dubai Economic Agenda, D33, sets out to double the size of the emirate’s economy by 2033. The Dubai Universal Blueprint for Artificial Intelligence exists to deliver a specific slice of that: AED 100bn a year from the digital economy, and a 50 percent increase in productivity.

These are published targets with named instruments behind them, not lines in a speech. The productivity figure is the one that matters for this page, because productivity is not a thing a government can produce on its own. It is produced inside companies, and most of Dubai’s companies are small.

The public half is being delivered

The part of the agenda that government controls directly is visibly moving. Chief AI Officers have been appointed across government entities. Tens of thousands of public employees are being retrained. AI commercial licences have been introduced. Land for data centres has been fast-tracked. The city is, in effect, running itself as a live experiment in what a government can do with this technology.

This is the part that can be directed. The instruments exist, the mandate is clear and the delivery can be seen. None of it involves us, and we cite it only because it is the context every owner in Dubai now operates in.

The private half is larger, and cannot be instructed

Small and medium businesses are around 95 percent of Dubai’s companies, 47 percent of its GDP and 52.4 percent of its workforce, according to Dubai SME and the Dubai Statistics Centre. A productivity target that depends on private companies can only be met inside them, one at a time, by people who understand how those companies actually operate.

Government cannot mandate what happens inside a privately owned company. It can set the direction, fund the training and publish the target, and it has. What it cannot do is sit in a twenty-person firm and move its quoting out of a spreadsheet. That work has no equivalent public mechanism, and it is the larger half of the agenda by every measure above.

The realistically addressable part is narrower than the 95 percent. It is the companies that are founder-led and small, already have customers and revenue, and are held back by their own operations rather than by demand. Those businesses can move quickly, because one person decides. They are also the ones with the most to gain, because the operation currently lives in that one person’s head.

The two-year transition: agentic AI across Dubai’s private sector

On 4 May 2026 the Crown Prince of Dubai, Sheikh Hamdan bin Mohammed, launched an initiative to move the emirate’s private sector to agentic AI within two years, announced through the Dubai Media Office. Gulf Business carried the key details of the plan. Two years from that date is May 2028.

The announcement gave the private half a mechanism for the first time. Dubai Chambers is to run training tracks across all of its business councils, reaching more than 14,000 companies with practical use of AI in daily business tasks. The same announcement set out incubators and funds alongside the training. Separately, du Business announced a programme to upskill 10,000 SMEs in agentic AI.

Training and funding are the right first instruments, and they reach a great many companies at once. What they cannot do, by their nature, is build the system inside any one of them. An owner can leave a training track knowing what agentic AI is and still go back to a business that runs on WhatsApp groups. The transition the announcement describes finishes inside each company or it does not finish. We have written separately about what the two-year window asks of a 20-person company.

The measurement problem

The UAE leads the world in AI adoption, at 70.1 percent of its working-age population in Microsoft’s AI Diffusion Report for the first quarter of 2026. The figure counts people using AI. It does not count businesses running on it. A founder who uses an AI assistant every day, while the company still runs on spreadsheets, group chats and personal memory, is inside that number.

Adoption is not transformation. A company has adopted AI when its people use it. A company has been transformed by it when the business itself runs on a system that holds its pipeline, its work, its money and its people, with AI carrying responsibility for outcomes inside that system. The first is widespread and is what the headline figure measures. The second is rare, and it is the only one that moves productivity at the level the Blueprint describes. We set out the distinction at length in Adoption is not transformation, and what agentic AI means in plain terms on its own page.

Inside a typical company

The tools are fine. Everything still routes through one person. That sentence describes almost every owner-run business we have sat down with, in Dubai and elsewhere, and the owner usually says it before we do. It looks like this:

• The post-job debrief that never happens.
• The client nobody brings back at six months.
• The documents you hunt for every time someone asks.
• A group chat per project, and no way to see across them.
• No idea which channel the revenue came from.
• Work that waits, because only you can unblock it.
• A relationship that went cold, found by accident.
• A renewal you forgot, that blocked a job.

None of these is a software problem in the usual sense. Each company already has software. These are the things that fall between the tools, and all of them land on the owner, because the owner is the only place the whole operation exists.

Our method

We map the operation, then we build the system that runs it. Four parts, the same every time. This is one product, the Symbaiotic OS, deployed company by company rather than invented afresh each time. What changes between deployments is the data and which functions a business needs, not the architecture.

The map comes first because the system has to be shaped around how the business already works, not the other way round. The executive comes last because it can only run a system that exists. How we work describes each part in turn, and the operating system page describes what the four parts hold.

A worked example: Mandala Creative Productions

Mandala is a creative production company with twenty nine years of history. It chose its Dubai entity to pilot its own AI transformation, because the market here moves fastest, and it chose us to build it.

We mapped the operation across a four hour session: the operators in the room, describing how the work really moves, one real production walked from first contact to final invoice. Every stage and every handoff was recorded in order, until the operation was out of one person’s head and on the wall. The Mandala OS is being built now, with the executive taking its first hats as each part goes live. The case study shows the room and the map.

Evidence

Systems in production, not prototypes. Two are live and one is in build, and the people who run those businesses use the result every day.

The method is not theoretical. All three are described, with what was mapped and what was built, under work in production.

What this leaves behind, beyond one company

Each deployment is an end in itself for the company concerned. It also leaves four things behind that matter to the agenda as a whole.

• A systematised business. Able to grow output without growing headcount, which is what a productivity target means at the level of one firm.
• A documented operation. Knowledge out of individual heads and into a system that survives them.
• A transferable method. The same four parts apply to the next company, and the next.
• A case study others can see. Which is how adoption spreads in a private sector that cannot be instructed.

The last point is the one we would underline. Owners copy owners. A documented case in their own sector, in their own city, does more to move a private company than a target does, and publishing those cases is part of the work.

We are building here, permanently

Symbaiotic is establishing its base in the UAE rather than serving it from elsewhere: the legal entity, a founder on the ground, and hiring locally as the work grows. The companies we work with are here, and the method only works in person. You cannot map an operation over a video call with the same result as a day in the room, and you cannot build a system around how a business runs without having watched it run. The two founders and how the work is split between them are described on their own page.

Working together

We are a small company and we are not asking for funding. Where we think we could be useful to the programmes, chambers and institutions carrying the two-year transition is in three places.

The useful things are access, a defined cohort, and permission to publish what works. The fuller version of this offer, written for programme managers, is on the programmes and partners page. Owners with questions of their own will find most of them answered under questions owners ask.

The target is set. It has to be built into the companies.

One business at a time, documented, so the next one is faster. That is the whole of our position on the agenda, and it is also what we do for a living. If you run a company in Dubai and recognise the description above, the conversation starts with a single session, at no cost, and you leave with our read on the real bottleneck either way.