A UK government pilot of Microsoft 365’s Copilot tool found employees were largely satisfied with the AI assistant, but the trial showed no evidence of productivity gains.
The study, conducted by the Department for Business and Trade, combined usage data, diaries, and interviews to measure Copilot’s impact during a pilot program.
Copilot tested across Word, Outlook, Teams, and other apps
The UK government’s Microsoft 365 Copilot trial ran from October to December 2024, issuing 1,000 licences across the Department for Business and Trade. About 70% of licences went to volunteers, with the rest randomly assigned to give a representative mix of roles and grades.
Copilot was embedded in Word, Outlook, Teams, Excel, PowerPoint, and a standalone app, with usage tracked through Microsoft dashboards. Staff diaries, interviews, and observed exercises fed into the evaluation, which was designed to examine time savings, user satisfaction, and adherence to acceptable use policies.
Strong satisfaction ratings
A majority of the Microsoft 365 Copilot users backed the tool, with 72% saying they were satisfied or very satisfied, and fewer than 1% voicing dissatisfaction.
The study highlighted other positive outcomes:
- 80% found Copilot useful in daily work.
- 60% reported a boost to overall job satisfaction.
- Written tasks such as drafting reports or summarising emails saw 74% to 80% satisfaction.
- Time saved per task: +1.3 hours drafting, +0.8 hours summarising research, +0.7 hours meeting notes.
There were more substantial benefits for some groups. Neurodiverse staff were statistically more likely to report higher satisfaction and recommend Copilot, while interviews showed that users with hearing or vision difficulties valued the tool for easing meetings and document work. Non-native English speakers cited better communication, wellbeing, and career confidence.
Extra work cancels out time savings
While staff saved time on core writing tasks, the evaluation found those gains were often offset elsewhere. Scheduling tasks, per the respondents, took around 35 minutes longer with Copilot, and generating images added about 30 minutes to each job.
Accuracy checks also slowed work. Nearly a third of staff said presentations and code reviews required more verification when Copilot was used.
In addition, interviews with colleagues outside the pilot found no visible change in output, consistent with the evaluation’s conclusion that satisfaction rose while productivity did not.
Productivity promise still unproven
The UK’s Microsoft Copilot study adds to evidence that AI’s workplace benefits remain uncertain. It aligns with other research showing the technology has no significant impact on earnings or hours, and with findings that some AI tools can even slow down productivity.
Industry leaders continue to argue the opposite. Microsoft’s Satya Nadella, for instance, has said up to 30% of the company’s code is now produced with AI tools, and Salesforce recently cut 4,000 jobs, citing AI’s ability to replace staff.
The clash between research and rhetoric leaves one question unanswered: Is AI truly delivering productivity, or just changing how work gets done?
In Washington, President Trump rolled out the red carpet for Altman, Cook, and Nadella, pitching his administration as a partner to AI’s biggest players.