Corporate report

Evaluating the Impact of Microsoft Copilot in HMRC

Published 9 July 2026

What is Microsoft Copilot?  

Microsoft 365 Copilot (hereafter referred to as Copilot) is Microsoft’s licensed AI-powered assistant. It uses large language models to understand prompts and generate responses, automate routine actions, and provide integrated contextual help across Microsoft (MS) apps like Word, Outlook, and Teams, as well as on the web and within its own dedicated app.   

Copilot Trials in HMRC

This report presents a summary of the findings from one phase of an ongoing evaluation of Copilot in HM Revenue and Customers (HMRC) and the Valuation Office Agency (VOA). The VOA merged with HMRC on 1 April 2026.

The HMRC and VOA Microsoft Copilot trial was conducted in three phases, with the third phase being the most comprehensive.

Two initial phases of testing were completed during mid-2024. These included 300 members of staff who were self-selected. The first trial ran for one month and the second for three months. Colleagues were asked to complete use-case and post-trial surveys asking questions about user satisfaction and time savings.

Phase III was the most comprehensive, running from September to December 2024 and evaluating the value of staff having access to a full Copilot license integrated with MS Office tools. In this phase 3,000 HMRC employees were randomly allocated a full Copilot license proportionate to business group distributions.

Early results from HMRC also informed the report from Government Digital Service (GDS), who led a wider cross-government trial involving 11 participating departments. 

Methodology of the Microsoft Copilot Phase III Trial

The Phase III trial used a mixed-methods approach, drawing on quantitative and qualitative data to provide a rounded assessment of Copilot’s impact. 

1. Trial Design and Participants:

• in the Phase III trial, 3,000 licenses were allocated randomly to colleagues in departmental roles that tended to predominantly use MS Office products in their day-to-day work
• an additional 500 licenses were directly allocated to selected employees who volunteered to trial Copilot capability for use as a reasonable adjustment or who had previously volunteered for earlier phases of the trials

2. Data Collection Methods:

• the primary data source was a survey completed by 1,364 users (45% of those colleagues randomly allocated a license)
• additional data was collected through usage monitoring statistics from the Microsoft Viva Insights dashboard, focus groups with 35 participants, and a task-based exercise with 49 participants

Results of the Microsoft Copilot Phase III Trial

The trial results were largely positive, showing strong staff uptake, satisfaction, and time saving benefits, while also identifying areas for further training and guidance.

1. Staff Satisfaction:

• 83% of staff allocated a license made use of it during the trial period
• the average satisfaction score was 7.1 out of 10, with 68% of users agreeing that Copilot was easy to use and integrated well into their workflow
• 61% of users reported that they would be disappointed to lose their license

2. Time Savings:

• the Government Digital Service cross‑government trial of MS Copilot (published June 2025), involving over 20,000 civil servants, reported average time‑savings of around 26 minutes per day, with more than 70% of users agreeing that Copilot increased their personal productivity or improved their ability to focus on more strategic-level tasks
• HMRC’s own controlled evaluation reported more conservative time-saving productivity gains, reflecting a deliberately cautious methodology, including a randomised cohort design, strict measurement standards and generally more diverse operational roles than those included in the wider GDS trial
• on average, HMRC participants self-reported saving 2-3% of their working week, equating to around 60 minutes per week for a full-time equivalent working pattern
• at the time of writing, analysts estimate that deploying up to 50,000 Copilot licenses until March 28, would generate around £50m of capacity generating, net productivity benefits per year
• reported time savings were consistent across various HMRC business groups and professions
• self-reported time-saving estimates were reduced by approximately 20% to reflect that not all license holders used Copilot and that survey respondents were more likely to be active users. The figures therefore should be treated as a baseline estimate rather than a precise measure of impact 

3. Usage and Benefits: 

• the most widely used functions were drafting documents in Word (53%) and collaborating on Teams (45%)
• users found Copilot particularly useful for summarising meetings, searching for documentation, and specific technical tasks such as writing code and Excel formulae
• the majority of users reinvested time savings to complete more ‘business as usual’ work, increase the quality of their work, and improve their wellbeing
• 64% of users reported that their use of Copilot increased over the course of the trial

4. Challenges and Concerns:

• survey responses showed that 46% of non-users selected security and data privacy concerns as the main reason for not using their license
• at the time of the trial, participants could only use Copilot on material below Official Sensitive classification  
• other challenges included the need for more practical and tailored training and time to learn how to effectively use Copilot