What problem does it solve?
Organizations buy assistant licenses for thousands of employees and then find that usage is uneven and the benefit is hard to prove. Staff get a tool that drafts, summarizes and searches across their mail, files and meetings, but no one decides which tasks it should handle, how outputs are checked or what to measure.
The public evaluations show the pattern. In the UK Department for Business and Trade pilot, around 64% of license holders used Copilot at least once a week, small time savings appeared on most tasks and the evaluation found no evidence that they led to improved productivity. The Australian Treasury found usage lower than expected and expectations that were not met. Microsoft reports 93% daily usage at Lloyds Banking Group, which made departments bid for licenses with measurable goals and built a network of volunteer coaches.
Every deployment on this page is a Microsoft 365 Copilot rollout, because those are the rollouts with published evaluations or reported figures, and two of the sources are Microsoft customer stories. The rollout questions (who gets a license, which tasks, what to measure) are the same for other office suite assistants.
How does it work?
- Choose the groups and the tasks. Pick departments with a business case and a measurable goal, and name the tasks the assistant should handle, such as meeting summaries, first drafts and document search.
- Prepare the data. Fix permissions and sensitivity labels on the files the assistant can reach, because it answers from whatever the employee can already open.
- License and train. Assign licenses, give role specific training and recruit local champions who coach colleagues.
- Measure use and outcome. Track weekly active users from the usage dashboard, run a diary study or time sampling for tasks, and compare with a control group where possible.
- Check outputs and adjust. Set rules for human review of anything external or consequential, then reallocate licenses from people who do not use them.
- Audience
- Employee facing
- Autonomy
- Copilot
- Adoption
- Mainstream
- Channels
- Microsoft Teams, Internal tools
What is it worth?
Benchmarks are computed from the public deployments below: one data point per organization per KPI, with who made each claim.
| KPI | Median | Reported range | Data points | Claimed by |
|---|---|---|---|---|
| Employee adoption | 90% | 64% to 93% Not pooled: up to 71% | 3plus 1 up to | 2 organization, 1 vendor |
| Hours saved | Not pooled | Not pooled: up to 2200 hours | 0plus 1 up to | 1 organization |
Value drivers: Employee productivity, Speed and cycle time.
Indicative value
A company with 5,000 office employees where 3,000 hold an assistant license
USD 495,000 to USD 6.3 million
Value of time saved per year
How this is calculated
Formula: licenses * activeShare * hoursPerWeek * weeks * rate. The low scenario uses every low input, the high scenario every high input.
| Input | Low | High | Basis |
|---|---|---|---|
| Licensed employees licenses, employees | 3,000 | 3,000 | The reference company. |
| Share of license holders using it weekly activeShare, fraction of license holders | 0.5 | 0.8 | Conservative against the benchmarks on this page: around 64% weekly use at the UK Department for Business and Trade (its own evaluation), 90% active use at Vattenfall (its annual report) and 93% daily use at Lloyds Banking Group (Microsoft's figure), while the Australian Treasury found most participants used it two or three times a week or less. |
| Hours saved per active user per week hoursPerWeek, hours per week | 0.25 | 1 | Editorial assumption. The DBT evaluation found small time savings, and UK government trial participants estimated 26 minutes a day, a self reported figure. Measure your own. |
| Working weeks per year weeks, weeks | 44 | 44 | Editorial assumption. |
| Cost per working hour rate, USD per hour | 30 | 60 | Editorial assumption for a loaded office staff cost, replace with your own. |
What it leaves out: Gross value of time, which only becomes a saving if the time is redeployed. It leaves out license cost, training, review of outputs and the DBT finding that time savings did not show up as improved productivity.
Who already uses it?
6 public deployments, strongest evidence first. Grades: A regulator or audit, B the organization itself, C vendor case study, D anonymous or estimate.
Vattenfall
Sweden · Energy and utilities · 2025
Vattenfall, the Swedish energy company, rolled out 6,000 Copilot licenses to employees and reports in its 2025 annual and sustainability report that 90 percent of licensed employees actively use the tool. The same report describes language model systems under development that let engineers get answers about earlier service and maintenance on hydro power plants. The report does not disclose a time saving or other outcome from Copilot.
- Employee adoption: 90%, licensed employees who actively use Copilot, 2025 annual report
"Vattenfall has successfully rolled out 6,000 Copilot licenses, with 90 percent of licensed employees actively using the tool – an impressive adoption rate."
Claimed by: organization
Department for Business and Trade
United Kingdom · Government and public sector · 2024
The UK Department for Business and Trade gave 1,000 Microsoft 365 Copilot licenses to staff for a three month pilot from October to December 2024 and published its own evaluation in August 2025. Around 64% of license holders used Copilot at least once a week. The evaluation found small time savings on most tasks, largest on written work, extra time on others such as scheduling and image generation, and no evidence that time savings led to improved productivity. It also reported hallucinations and uneven quality checks.
- Employee adoption: about 64%, license holders using Copilot at least once a week, October to December 2024
"This data implies around 64% of licence holders used M365 Copilot at least once a week, while around 30% used it at least once per working day."
Claimed by: organization
Department of the Treasury (Australia)
Australia · Government and public sector · 2024
The Australian Department of the Treasury ran a 14 week trial of Microsoft 365 Copilot from 20 May to 23 August 2024 with 218 staff, as part of the whole of government trial coordinated by the Digital Transformation Agency. The Australian Centre for Evaluation found a smooth technical implementation but usage lower than expected, with most participants using Copilot two or three times a week or less, and expectations that were not met. The evaluation did not measure time saved directly.
No outcome disclosed.
Government Digital Service
United Kingdom · Government and public sector · 2024
The Government Digital Service ran a trial of Microsoft 365 Copilot with 20,000 employees across UK government from 30 September to 31 December 2024. Copilot in Teams was the most used application throughout, with adoption peaking at 71%, and one participant named summarizing meeting notes among the tasks where it helped. Participants estimated an average saving of 26 minutes a day across all tasks; that figure is self reported, covers every Copilot use and is not recorded as a meeting metric. The report also notes weaker results on complex, nuanced or context heavy work, and concerns that Copilot relied on external sources without built in verification.
- Employee adoption: up to 71%, peak share of trial users using Copilot in Teams, October to December 2024
"Teams was the most popular tool for M365 Copilot and remained dominant throughout the experiment with a maximum adoption of 71%."
Claimed by: organization
Lloyds Banking Group
United Kingdom · Banking · 2025
Lloyds Banking Group started with 300 Microsoft 365 Copilot licenses in an early access program, let each department bid for licenses with a business case and measurable goals, and grew to 30,000 licenses, with a commitment in March 2025 to roll out 40,000. Microsoft's customer story reports that 93% of licensed colleagues use Copilot daily. The bank trained more than 10,000 people and built a network of about 1,000 volunteer coaches. Colleagues use it to draft and summarize emails and documents and to prepare for and record meetings.
- Employee adoption: 93%, daily Copilot usage among colleagues with licenses
"More than 10,000 people have now had training, and this has contributed to a 93% daily Copilot usage rate among the 30,000 colleagues with Copilot licenses."
Claimed by: vendor
Ma'aden
Saudi Arabia · Energy and utilities · 2025
Ma'aden, the largest multi commodity mining and metals company in the Middle East, gave employees Microsoft 365 Copilot and built two Copilot Studio agents inside Microsoft Teams: one for governance documents and one for delegation of authority policies. Employees use Copilot to draft emails and presentations, summarize content, take meeting minutes and define actions, and pull figures out of Excel files. Ma'aden started with early adopter champions in each business unit, trained its executives first, and reports the time saved in the first phase of the rollout.
- Hours saved: up to 2200 hours, per month, first phase of the rollout
"We have more than 36,600 interactions using Microsoft 365 Copilot and collectively save up to 2,200 hours in a month in the first phase alone"
Claimed by: organization
How do you implement it?
A model agnostic playbook: what to prepare, the order to build in, and what goes wrong.
Data you need
- Current permissions and sensitivity labels on mail, files and sites
- A list of target roles and the tasks the assistant should handle
- A baseline of the time those tasks take today
Systems to integrate
- Office suite and collaboration platform
- Identity and access management
- Usage analytics and license management
- Data loss prevention and sensitivity labelling
Complexity: Medium
The software is bought, not built. The effort sits in cleaning up file permissions and labels, training, review rules for external content, and a measurement plan that goes beyond usage counts.
- 1
Pilot with a mix of volunteers and random users
The DBT pilot allocated about 70% of licenses to volunteers and 30% to a random sample, so the findings were less skewed toward enthusiasts.
- 2
Fix access before launch
Review oversharing on files and sites. The assistant shows an employee what they are allowed to see, including things they never knew they could open.
- 3
Train by role
The DBT evaluation found self led training raised satisfaction more than formal sessions. Lloyds trained more than 10,000 people and built a network of volunteer coaches.
- 4
Measure beyond usage
Combine dashboard usage with task level time measurement and a control group, and separate the time saved from the productivity gained.
- 5
Reallocate licenses
Review active use quarterly and move licenses from inactive holders to teams with a case.
Guardrails
- Sensitivity labels and permissions reviewed before licenses are issued
- Human review of any output sent outside the organization
- Clear rules on what data may be entered and which tasks are out of scope
- Logging and retention of assistant use for audit and legal hold
KPIs to instrument
- Weekly active share of license holders
- Time per task with and without the assistant, from observation or diaries
- Reported hallucinations and corrections
- Licenses reassigned per quarter
Human in the loop
The employee owns every output. Managers review high stakes documents as before, and the program owner reviews usage and quality data every quarter.
Common failure modes
- Licenses without use
- At the Australian Treasury most participants used Copilot two or three times a week or less. Measure weekly use and reassign idle licenses.
- Time saved, no productivity gained
- The DBT evaluation found no evidence that time savings improved productivity. Decide in advance where saved time goes.
- Output quality
- Evaluations record hallucinations and uneven quality on slides and scheduling. Set review rules and exclude weak tasks.
- Oversharing
- The assistant exposes files with loose permissions. Clean up access first.
What are the risks and rules?
EU AI Act
Depends on design
Drafting, summarizing and searching for employees is not an Annex III use. Article 50 puts the marking of generated content on the provider; the organization as deployer must disclose AI generated text only when it publishes it to inform the public on matters of public interest without human review or editorial responsibility (Article 50(4)), and Article 4 requires AI literacy for the staff who use it. It becomes high risk under Annex III point 4 when the organization uses it to recruit, evaluate or monitor employees, for example to screen candidates or draft performance ratings, so the organization must set and enforce the allowed tasks.
Guidance
- Microsoft 365 Copilot Experiment, cross government findings report (Government Digital Service, Europe). Findings from 20,000 UK government employees, including weaker results on complex work and concerns about unverified external sources.
Controls to put in place
- License register with owner, purpose and review date
- Data labelling and access review before rollout
- Usage and quality reporting each quarter
- Staff guidance on acceptable use
Frequently asked questions
- How many employees actually use the assistant after a rollout?
The UK Department for Business and Trade reports that around 64% of license holders used Copilot at least once a week in a three month pilot, and Microsoft reports 93% daily usage among the 30,000 licensed colleagues at Lloyds Banking Group. The Australian Treasury found most participants used it two or three times a week or less.
- Does the assistant improve productivity?
The evidence is mixed. The DBT evaluation found small time savings, largest on written tasks, and no evidence that they led to improved productivity. In Microsoft's customer story, Ma'aden reports up to 2,200 hours saved in a month in its first phase, and UK government trial participants estimated an average of 26 minutes saved a day, a self reported figure.
- What should be done before issuing licenses?
Review file permissions and sensitivity labels, choose roles and tasks, plan training and decide how to measure results. The assistant answers from whatever the employee can already open.
How to cite this page
Blits.ai AI Use Case Library, "Workplace AI assistant rollout across an organization", last verified 12 November 2026, https://www.blits.ai/ai-use-cases/workplace-ai-assistant-rollout. Licensed under CC BY 4.0. Method: how we verify use cases.
Changelog
- 12 November 2026: First published