On 4 May 2026 the Crown Prince of Dubai launched the move of the emirate’s private sector to agentic AI — two years, through the Dubai Chamber’s business councils, with incubators, funds and training tracks behind it. The hard part was never the ambition. It is standing up agents that an auditor, a regulator and your own board can all interrogate. That is the entire design of what we build.
Photo: Unsplash
His Highness Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum, Crown Prince of Dubai, launched the transition of Dubai’s private sector to agentic AI — self-executing, self-leading systems — over a two-year horizon.
Every organisation that moves fast here will reach the same room: a regulator, an auditor or a board member asking where a number came from, which rule applied, and who decided. “The model produced it” ends that conversation badly. So we build agents where the model is structurally incapable of producing the number in the first place.
Loads the data. Computes every figure. Resolves the rules in force on the date being judged. Decides whether the run may proceed at all. If a rule blocks, the model is never called.
Shown the names of the figures, never their values. Every number it writes is substituted from the calculation; a figure it invents fails the run. No free-text prompt from a person ever reaches it — a guard test fails the build if one does.
Most agentic-AI proposals staff one department and call it a transformation. A two-year shift has to reach the executive floor and the back office in the same programme, or the two halves end up reasoning from different facts.
Continuous intelligence across People, Process, Market, Technology, Finance and Offer. It names the binding constraint before anyone proposes a solution to it, and decides what the business builds next. This is the layer that answers to the board.
Xamun Intelligence →A roster of agents doing the routine work of named roles — the daily performance pack, the weekly management report, the monthly board pack, the case-readiness check, the campaign compliance gate, the vendor SLA reconciliation. Reports arrive as a PowerPoint pack in the inbox, each deck ending on a provenance slide. People are interrupted at judgement forks and nowhere else, and the hours returned are measured against a declared baseline.
WorkForceOS →Twenty-five production systems already built and running — escrow, tenancy, brokerage, land, towage, bunker fuel, wealth, lending, clinics, insurance. If one fits your operation you are not starting from a blank page; you are configuring a system that already passed its own compliance surface.
All twenty-five systems →The thing that makes you difficult to compete with is, by definition, the thing nobody sells. The Xamun Lights-Off Software Factory builds it — a working, deployed system in thirty days or less, source available, running on your infrastructure and under your brand.
The Software Factory →What makes it one system rather than four purchases: Xamun Intelligence and WorkForceOS resolve their rules from the same GraphIQ knowledge base — so the law the executive layer reasons about is the law each agent actually applied, on the date it applied it. WorkForceOS reports upward in rows, not summaries: hours returned per role with the baseline each rests on, which decisions run unattended and on what evidence, every rule evaluation with its citation, and how often a person was needed and why. That last figure is the real map of where your organisation’s judgement is still required — and it is the one the intelligence layer uses to decide what to build next.
Several of the twenty-five OS Series systems were built against UAE regulation from the first commit — not ported to it afterwards. If one covers your operation, the two years look considerably less tight.
The loudest claim any workforce product makes is the number of hours it gives back. So the ledger is built to survive the question that follows it — and that means measuring the before, honestly, while your people work exactly as they always have.
Xamun Intelligence reads across People, Process, Market, Technology, Finance and Offer, and names the one bottleneck worth moving. Not a list of twelve opportunities.
Agents run alongside your team and propose; nobody acts on them yet. A named person declares each baseline with its basis — a time study, a manager’s estimate — and it goes on the ledger with their name against it.
The demonstration data is swapped for your warehouse, your service desk, your contract register. No agent is touched to do it. The GraphIQ pack is curated to your jurisdiction and industry.
Run at approve. Sign, mark what was right and wrong, and let the record recommend each promotion. Anything customer- or regulator-facing stays with a person, permanently and by design.
His Highness Sheikh Hamdan bin Mohammed bin Rashid Al Maktoum, Crown Prince of Dubai, launched an initiative to move Dubai's private sector to agentic AI — self-executing, self-leading systems — over two years. It runs through the business councils affiliated with the Dubai Chamber of Commerce, and is backed by incubators for agentic-AI companies, dedicated funds, and specialised training tracks. The stated goal is for Dubai to become the world's leading city in adopting these technologies economically and commercially.
If you are a member of a Dubai Chamber business council, the training tracks and funds are pointed at you directly. Beyond that, the practical pressure is commercial rather than statutory: your customers, your bank and the government entities you transact with are all moving on the same clock. Xamun's read is that the binding question is not whether to adopt agentic AI, but whether what you adopt can survive an audit.
A chatbot waits for a person to type. An agent serves a named role, runs on a schedule or when a record arrives, and produces work. The distinction that matters to a regulator is provenance: a chatbot cannot tell you where a number came from, which rule applied, or who decided. In WorkForceOS every run is a row — the calculation, the rules applied with their citations, what the model was shown, and who resolved what.
No, and that is deliberate. The reporting agents produce a PowerPoint pack and email it — a short brief in the body, the deck attached — to a distribution list named per report. The people who read a board pack are rarely the people who work in the system, so they are named in a directory rather than given accounts. A system that makes an executive log into a dashboard to find their own report has moved the work, not removed it. Each deck ends on a provenance slide: the run, the obligations pack and version, how many rules applied and how many blocked, and which model drafted the prose alongside the statement that no figure in the deck came from it.
Xamun Intelligence sits on the executive floor: it reasons across People, Process, Market, Technology, Finance and Offer, names the binding constraint, and decides what the business should do next. WorkForceOS sits beneath it and does the routine work of every other named role — the daily pack, the weekly report, the compliance gate, the vendor reconciliation — and reports back in rows the intelligence layer can read. Both resolve their rules from the same GraphIQ knowledge base, so the law the executive layer reasons about is the law each agent actually applied, on the date it applied it.
Every rule an agent applies carries its instrument, its article and its pack version, and every run records whether the knowledge base answered live or from an offline copy. A run is judged by the law in force on the date of the event. Anything customer- or regulator-facing is approved by a named person every time and cannot be made autonomous — the server refuses it. No customer identifiers leave for the knowledge base: it receives amounts, flags and risk tiers, never names, Emirates ID numbers or nationality. This is compliance support, not a legal opinion.
Thirty days or less through the Xamun Lights-Off Software Factory — a working, deployed system, not a prototype and not a roadmap. The OS Series exists because twenty-five of those builds are already finished; if one of them fits your industry, the clock is shorter still.
Under Outcome-as-a-Service there is no upfront CAPEX. Xamun invests the build and is paid on what the system actually processes. A pilot begins with two to four weeks of baselining — agents run in suggest mode while your people work as they always have — so the hours returned are measured against a declared baseline rather than claimed in a slide.
Bring us the routine work of one role. We will show you a governed agent doing it — with the rule it was judged by, the citation beneath it, and the hours measured against a baseline you declared yourself.