The headline figures are good for a rollout of this size. They also come with several caveats that matter if you run, or are about to run, a Copilot deployment.
The numbers MCIT presented
The indicators showed active users above 8,000, an adoption rate of 69.6 percent among 12,260 licensed users, and roughly 2.1 million actions and tasks recorded with the tool.
The training figures were:
- 508 training sessions delivered to more than 9,000 employees.
- Session types that included foundational training, executive leadership sessions, use-case discovery workshops, specialized training and skills development for the teams supporting adoption.
- 96 AI agent use cases under the program, with selected practical demonstrations at the meeting.
The program's support mechanisms were communications campaigns, activation activities, learning resources, dashboards and periodic reviews. MCIT also honoured entities for excellence in digital transformation adoption. The report doesn't name the winners or give the criteria.
The ministry's stated priorities for the next phase are goals, not results. Assistant Undersecretary Sami Mohammed Al Shammari said the results reflect clear progress. He said the focus now moves to expanding usage, developing higher-value use cases and strengthening institutional impact.
Reading the math
Multiplying 12,260 licences by 69.6 percent gives roughly 8,500 users. That fits "more than 8,000". The report doesn't give an exact active-user count, so treat my figure as arithmetic and not an official number.
Three details are missing from the report:
- Measurement window. There's no stated period for "active", so it could mean active in the last 30 days or at any point in the program.
- What counts as an action or task. The 2.1 million figure has no definition and no time interval. It also isn't stated whether it counts unique work items.
- Entity breakdown. Nothing shows whether a few large agencies are carrying the average.
For comparison, a UK Government Digital Service trial of Microsoft 365 Copilot, a separate study in another country, defined its terms explicitly. It ran from 30 September to 31 December 2024 with 20,000 government employees, according to the GOV.UK report. That report counted an active user as someone with at least one interaction in the previous 30 days. It defined adoption as active users divided by licensed users. I can't say Qatar used the same definition, but this is the kind of detail that makes adoption percentages comparable between programs.
How it compares with phase one
MCIT's December 2025 account of the first phase gave these results: 62 percent adoption, more than 9,000 active users and 1.7 million tasks, which MCIT said saved more than 240,000 working hours. That first batch involved nine governmental and semi-governmental entities. The program was then expanded to 17 entities, with specialised training through the Qatar Digital Academy.
The new report shows a higher adoption rate (69.6 percent against 62 percent) and more tasks (about 2.1 million against 1.7 million). Active users are lower, at more than 8,000 against more than 9,000. I wouldn't read this as a clean trend, because:
- The participant groups differ.
- The reports don't state matching measurement windows.
- The wording differs. The MCIT site describes the first-phase users as daily active users, while the 2026 report says only "active users".
The 240,000 hours-saved claim belongs to phase one. The 2026 report makes no productivity or time-saved claim for the second cohort, and none of these figures has been independently audited.
Why agents are the interesting part
The report describes a move from individual use of AI tools to applications tied to institutional priorities and knowledge. That is the 96 agent use cases. The report doesn't say how many are built, in production, or used by staff. "Use case" here may mean anything from an idea to a deployed agent.
The UK trial offers a caution on agents. It found that participants' early experiments with Copilot agents struggled to identify the exact documents used to generate responses when using a OneDrive folder as the source. That isn't a finding about Qatar. It is a reason to test source attribution early when an agent is meant to draw on institutional knowledge.
Governance questions for IT admins
Copilot works within each user's existing permissions. The GOV.UK report explains that the tool adopts the permissions of the end user and only retrieves documents the user could normally access. The consequence is that existing oversharing in SharePoint, OneDrive or Teams becomes easier to find. The UK report said organisations should keep information management current. It also said most trial organisations turned off internet access to rely on internal sources.
This is general guidance, not a finding about Qatar's entities. For any large Copilot rollout, a practical checklist looks like this:
- Audit file and site permissions before licences go out.
- Decide whether web grounding is enabled.
- Define what "active user" means and record the measurement window before you report adoption.
- Pair licences with role-based training, as MCIT's 508 sessions suggest.
- Test agents for source attribution before widening access.
Bottom line
On the reported numbers, Qatar's programme is scaling. Nearly 70 percent of 12,260 licences are active, and training has reached more than 9,000 people. The open questions are the ones any Copilot buyer should ask. What counts as active, over what period? Did the tasks produce better services? Do the agents hold up in production? MCIT hasn't answered those publicly yet.