A UK government trial of 20K+ civil servants using Microsoft Copilot for three months found that they saved 26 minutes per day on average, or two weeks per year
Staff used Microsoft's Copilot to help draft documents and summarise meetings — Simply sign up to the Artificial intelligence myFT Digest …
Context & Ripple Effects
Microsoft had already positioned Copilot as an Office-integrated paid assistant, while early Microsoft 365 testers questioned its $30-per-seat value even as they found it more useful in Teams than Excel. This large public-sector trial adds a concrete productivity signal for the drafting and meeting-summary tasks where the product is being used.
The result also sits ahead of much larger enterprise deployments, including Accenture's planned Copilot 365 rollout to roughly 743,000 employees. It matters because it shifts the debate from feature access toward whether time saved in routine knowledge work can support broad-seat adoption.
First-order effects
- For participating civil servants, Copilot changes the immediate workflow for document drafting and meeting summarization, with reported average time savings of 26 minutes a day over the three-month trial.
- Microsoft gains a sizable institutional reference point for Copilot's workplace value proposition, particularly for the collaboration-oriented uses that early Microsoft 365 testers found strongest in Teams.
Second-order effects
- Large employers and public-sector buyers can use measured time savings—not just model capability—as a procurement benchmark, raising pressure on vendors to demonstrate value by task and role.
- The finding sharpens scrutiny of Copilot pricing: time saved may support wider licensing in suitable workflows, but the earlier concerns over the $30-per-seat price suggest average productivity claims will still need to translate into budget-specific returns.
Third-order effects
- If similar results hold across organizations, office-suite AI is likely to be evaluated as a layer of everyday work infrastructure, with adoption increasingly tied to workflow measurement and governance rather than standalone chatbot usage.
- The broader market may consolidate around assistants embedded in the documents, meetings, and collaboration systems where work already happens; the durability of that shift depends on whether gains persist beyond trials and across different job types.
The trend: This is one data point in the move from selling workplace AI on model novelty to proving its cost per useful task inside established productivity suites.