UK Government Digital Service
UK government trial gives 20,000 civil servants Microsoft 365 Copilot and finds 26 minutes saved a day
The UK Government Digital Service ran a three-month trial of Microsoft 365 Copilot with 20,000 staff in 12 organisations, including DWP, HMRC and the Home Office. Participants reported saving an average of 26 minutes a day, mostly on drafting and admin. The government stresses the figures are self-reported.
Key figures
Copied exactly from the sources. Copy any figure with its citation.
82%
Participants who said they would not want to return to working without Copilot2
83%
Adoption rate (users active in the previous 30 days) after change management in October 20242
17%
Participants who noticed no clear time savings2
What it took
The problem
Government departments spend large amounts of staff time on routine work such as drafting documents, summarising emails and updating records. The UK government wanted evidence on whether a commercial AI assistant could cut that time before deciding on wider use.
What they built
The Government Digital Service (GDS) ran a cross-government trial of Microsoft 365 Copilot. This is an assistant built into Word, Excel, PowerPoint, Outlook and Teams, with a chat tool, which drafts and summarises content and searches the user's own files. Staff at 12 organisations took part, among them the Department for Work and Pensions, HMRC, the Home Office, the Ministry of Justice and Companies House.
How they did it
GDS gave 20,000 employees a licence, with each organisation committing to at least 1,000 licences. The trial ran for three months, from 30 September 2024 to 31 December 2024. Most organisations switched off internet search so the tool relied on internal data. GDS collected usage data, a survey with 7,115 responses, feedback forms and five focus groups. Change management and licence rollout in October lifted adoption to 83%, and it stayed near 80% afterwards.
Results
Participants reported an average saving of 26 minutes a day. The press release says this adds up to nearly two weeks per person per year. The report adds that 82% would not want to go back to working without the tool, and that 17% noticed no clear time savings. Drafting documents brought the largest task-level savings. Benefits were smaller where work needed nuance or involved sensitive data, and some users were cautious about security.
What to copy
Plan training and communication from day one, and measure by role, not only on average.
Lessons for your project
Budget for change management, not only licences. The report links the roughly 80% active use to engagement work and training.
Expect uneven benefits by role. The roles with the lowest satisfaction saved the least time, and the report cites policy-focused teams struggling with nuance as an example.
Fix document access rules before rollout. Copilot searches what each user can open, which can expose files people should not reach.
Keep human review in place. The report says oversight was needed at all times.
Caveats
- The time savings are self-reported by participants in a survey, not measured. The government says the figures are averaged across the full cohort of 20,000.
- The report says it could not identify how the saved time was spent.
- The trial ran over the festive period, and the report says this shortened effective use and may have lowered adoption.
- Both sources are published by the UK government, which ran the trial. Microsoft supplied the tool and is quoted in the press release.
- The sources name the product, Microsoft 365 Copilot, but not the underlying language model.
- The deployed date is the trial start date, not a permanent rollout.
Sources
- Landmark government trial shows AI could save civil servants nearly 2 weeks a year
GOV.UK (Department for Science, Innovation and Technology) · 2025-06-02 · Press release · gov.uk
- Microsoft 365 Copilot Experiment: Cross-Government Findings Report
GOV.UK (Government Digital Service) · 2025-06 · Research paper · gov.uk
Summarized in our own words from the sources above. Spotted an error? See corrections.