Analysis
Abu Dhabi's AiNative: what AI training for 20,000 government staff means for enterprises
1AYM is an OpenAI Select Partner.
Published . Vendor facts checked on against each vendor’s own pages, listed at the end. Plans, prices and availability change. Neither OpenAI nor Anthropic has reviewed this analysis.
The short version
The sections below take the target, the curriculum and tools, and the public programme in turn, each with what to check before you plan your own.
- The target: 20,000 staff by year end
- A headcount and a date, announced on 7 October 2026, after Microsoft 365 Copilot went to 35,000 government employees. A pilot inside government is already running. My take: a headline number, and the real prize is domain experts using AI.
- The curriculum and the tools
- Foundation courses, an executive programme and later certifications in coding and data, on ChatGPT, Codex and Microsoft 365 Copilot. DGE owns the curriculum and calls it technology-agnostic.
- The public programme
- Citizens and residents will get a course through DGE's Tomouh app, signing in with UAE Pass, delivered on the free version of ChatGPT. No opening date is given.
20,000 government employees by the end of 2026
What was announced
The Department of Government Enablement (DGE) announced AiNative at Ai Everything Abu Dhabi, in a release dated 7 October 2026. It describes an AI training programme for every government employee and for the wider public, built with OpenAI, Microsoft and the Mohamed bin Zayed University of Artificial Intelligence (MBZUAI). The stated aim is to upskill 20,000 government employees by the end of the year, as part of Abu Dhabi's drive to become what it calls the world's first AI-native government by 2027.
The programme is being rolled out in phases, and an initial pilot within government is already under way. DGE says it follows the rollout of Microsoft 365 Copilot to 35,000 Abu Dhabi Government employees through the Frontier Employee Programme, announced in July 2026.
What it means for enterprises
The order of events is the useful part. The licences came first, in July, and the training target followed three months later. An enterprise that does the first half and stops is left with a licence bill and usage that depends on who taught themselves. DGE's own framing is that tools and skills have to progress together.
A target with a number and a date is also something a programme can be held to. An AI adoption plan that says staff will be offered training cannot fail. One that says 20,000 people by the end of the year can, which is why it is more likely to be resourced.
For Gulf enterprises there is a second reading. Government entities are customers and regulators. When officials are trained to use AI in daily work, suppliers, banks and contractors that deal with them are likely to be asked how their own staff use it, and how that use is governed.
Before you switch it on
- A number and a date
- Set a headcount and a deadline for your own programme, and name the person who reports against it.
- Licences against skills
- Compare seats bought with people trained and with people active each week. The gap is the size of the job.
- A pilot first
- Run one department through the course before the whole organisation, as DGE is doing, and change the course on what you learn.
- What counts as trained
- Define it before you start: attended, assessed, or using the tools in real work a month later.
My take
I say it is a headline number. We have people who have done real enablement across large enterprises, so we know their needs and the challenges, especially with non-technical people who want to use their domain expertise with AI.
That is the superpower, and I do not think many have clocked on to it yet. For example, the system a large international marketing agency runs on Claude and Notion is a thing of beauty. Around a thousand people who are all experts at different things work together to make their company's context specialised. It is now far removed from the generic LLMs you get out of the box, and still nimble enough to switch vendors whenever it needs to.
Sources: [1]
The curriculum: ChatGPT, Codex and Microsoft 365 Copilot, owned by the government
What was announced
AiNative offers foundation learning journeys for government employees and for the public, and a dedicated executive programme for department leaders. The government journey teaches practical tools, named as ChatGPT, Codex and Microsoft 365 Copilot, with the aim that employees use AI confidently and responsibly in daily work. Specialised, function-specific certifications, in areas such as coding and data, will follow in later phases.
DGE owns the curriculum, which was shaped with OpenAI, Microsoft and MBZUAI. The release describes AiNative as open and technology-agnostic, with room to add further partners over time.
What it means for enterprises
Three design choices are worth copying. The curriculum belongs to the organisation, with vendors contributing to it. Leaders get their own programme, since a department head who has never used the tools cannot judge what their team should do with them. And role-specific certification comes after a common foundation, so that everyone shares a baseline before the specialists go deeper.
The vendor list tells you which tools Abu Dhabi's officials will be working in for now: OpenAI's and Microsoft's. The technology-agnostic wording tells you not to read it as permanent. A Gulf enterprise choosing a platform can take the signal without treating it as a mandate.
Vendor-led training teaches the product. What staff need on top is the organisation's own rules: which data may go into which tool, which outputs need a second person, and who to ask. Owning the curriculum is how that part gets in.
Before you switch it on
- Ownership
- Keep the curriculum yours. Use vendor material for the product and write the policy, data and judgement modules yourself.
- Leaders
- Give executives a session of their own, built on decisions they make, before their teams are trained.
- Roles
- Plan the second phase by function, such as finance, legal, engineering and service teams, each with its own examples.
- Portability
- Teach habits that survive a change of tool: checking sources, protecting data and knowing when to stop and ask.
Sources: [1]
A public course through the Tomouh app
What was announced
Citizens and residents will be able to take the public learning journey through Tomouh, DGE's AI-powered learning app, by logging in with UAE Pass. It will cover practical uses of AI in everyday life, including managing personal finances, running a small business, finding a job, studying and creating content responsibly online, and will be delivered through the free version of ChatGPT. The release says the programme will be made public after the government pilot, without giving a date.
What it means for enterprises
A public course changes the labour market an employer hires from. If it reaches scale, new joiners in the emirate will arrive having used an AI assistant for real tasks, and an employer's induction can start from there.
It may also mean more practised customers. A bank, insurer or retailer in the UAE can expect more customers to check an offer, a contract or a complaint response with an AI assistant before replying, and to arrive better prepared.
Before you switch it on
- Induction
- Ask new joiners what AI tools they already use, and teach the company's rules on top of that.
- Customer documents
- Assume your terms, letters and quotes will be read with an AI assistant, and check how they read when one summarises them.
- Small suppliers
- If you rely on small local suppliers, a public course is a cheap way to raise their baseline. Point them at it when it opens.
Sources: [1]
Where the market is heading: licences first, then skills at scale
The order in this announcement is worth noting. The tools were rolled out first, as a platform decision. The training follows as its own programme, with a budget, a target and an owner, because the licence alone did not change how people work.
The UK government's own guidance makes the same point for public bodies. The Artificial Intelligence Playbook for the UK Government, published on 10 February 2025, sets ten principles, and the ninth is that you have the skills and expertise needed to implement and use AI solutions. Abu Dhabi has put a number and a date on that principle.
For enterprises in the UK and the US, the comparison is uncomfortable in a useful way. An AI strategy and a set of licences are common. The test Abu Dhabi has set itself is harder: how many people have been trained, to what standard, and by when the rest will be.
What is not known yet
Everything above comes from the Abu Dhabi Media Office release of 7 October 2026. It is an announcement, and it leaves the delivery detail open.
- What upskilled means
- The release gives no course length, assessment or pass standard, so 20,000 could mean anything from an hour online to a certified course.
- Which licences staff hold
- It names ChatGPT, Codex and Microsoft 365 Copilot as tools taught. It does not say which ChatGPT plan government employees use, or how many have one.
- The public timetable
- There is no date for the public course, only that it follows the pilot.
- Results
- No measure of success beyond the headcount is published, such as usage after training or time saved.
Sources: [1]
For engineers: Codex in the curriculum, and what a coding track needs
Codex is named alongside the office tools
The government journey lists Codex with ChatGPT and Microsoft 365 Copilot, and coding is one of the function-specific certifications due in later phases. The release gives no detail on either. For an engineering leader it is still a signal: a coding agent is being taught as a general workplace tool, in the same course as the office assistants.
Sources: [1]
What a coding track needs that an office course does not
Training engineers on a coding agent is mostly about the environment around it. A course is worth little unless the repositories, permissions, review rules and CI checks the agent works inside are set up first, because that is where the habits form. Teach on the organisation's real setup, with its own instruction files and its own review standard, and measure the outcome in merged, reviewed changes.
Where 1AYM fits
Rolling out AI tools to a workforce is core 1AYM work: we roll out ChatGPT Enterprise and Claude Enterprise, we have rolled out Claude Enterprise for a client, and 1AYM is an OpenAI Select Partner. We work with clients across the UK, the US and the Gulf.
For an enterprise that wants its own version of this programme, the place to start is the target, the definition of trained and one pilot department, in that order. Where engineers are in scope, our AI Engineering Transformation work sets up the environments and checks the training depends on. It does not have to start big: 1AYM takes small fixed-scope statements of work as well as larger builds.
Sources
Frequently asked questions
What is AiNative?
An AI training programme launched by the Department of Government Enablement, Abu Dhabi, announced on 7 October 2026. It was built with OpenAI, Microsoft and the Mohamed bin Zayed University of Artificial Intelligence, and is for every government employee and, later, the wider public.
How many people will AiNative train, and by when?
The stated aim is to upskill 20,000 Abu Dhabi government employees by the end of 2026. An initial pilot within government is under way. The public course has no published date.
Which AI tools does AiNative teach?
The release names ChatGPT, Codex and Microsoft 365 Copilot for government employees, and the free version of ChatGPT for the public course. It describes the programme as technology-agnostic, with room to add partners.
What should an enterprise copy from it?
Four things: a headcount and a date, a curriculum the organisation owns, a separate programme for leaders, and a pilot before the full rollout. Add a definition of what trained means, which the announcement does not give.
Further
- ChatGPT Enterprise rollout · A 90-day plan for the licence side of the same job: identity, retention, a pilot and measurement.
- Microsoft 365 Copilot oversharing checklist · The permissions work to finish before staff are trained to search everything they can reach.
- Gulf AI data residency · What the law mandates, where the model runs and where inference happens, for buyers in the Gulf.
- AI implementation: the first 30 days · What the first month of an AI programme should deliver, week by week, including handover.
Planning your own programme?
Half an hour is enough to set a realistic target, decide what trained should mean and pick the pilot department.
Last reviewed · 1AYM