Shadow AI: what the surveys count, and what to do about it
Shadow AI is AI used for work outside the company's approval or oversight.
In Microsoft UK's Censuswide survey of 2,003 UK employees aged 18 and over in October 2025, 71 percent said they had used unapproved consumer AI tools at work. In Salesforce's YouGov survey of more than 14,000 full-time employees in 14 countries, conducted from 18 to 31 October 2023, over half of the 28 percent who used generative AI at work did so without their employer's formal approval.
Before choosing a response, find out what people are trying to get done and why they use those tools. A missing approved option calls for a different decision from uncertainty about the rules.
What is shadow AI?
Shadow AI is AI used for work without the company's approval or oversight.
A personal account or a tool paid for at home is shadow AI only when the company has not approved it for the work and the information involved. An employee who is unsure about permission is also in a different situation from someone who knowingly ignores a ban.
Shadow AI statistics: what each survey counted
One shadow AI figure is a share of all employees, another a share of AI users only, and they count different behaviour.
| Survey and fieldwork | Who was asked | What was counted | Result |
|---|---|---|---|
| Microsoft UK, October 2025 | Who was asked 2,003 UK employees aged 18 and over | What was counted Used unapproved consumer AI tools at work | Result 71 percent of all employees asked; 51 percent still do every week |
| KPMG and the University of Melbourne, November 2024 to mid-January 2025 | Who was asked 48,340 people in 47 countries, 32,352 of them employees | What was counted Used AI tools at work without knowing whether it is allowed | Result Employees reported use without knowing whether it was allowed; the report does not make clear whether non-users were included |
| Software AG, 13 to 25 September 2024 | Who was asked 6,000 knowledge workers aged 18 and over who mainly work at a desk or computer, in the US, UK and Germany | What was counted Uses personal or otherwise non-company issued AI tools | Result More than half of the 75 percent who use AI |
| Microsoft and LinkedIn, 15 February to 28 March 2024 | Who was asked 31,000 full-time employed or self-employed knowledge workers in 31 markets | What was counted Brings their own AI tools to work | Result 78 percent of the 75 percent who use AI at work |
| Salesforce, 18 to 31 October 2023 | Who was asked More than 14,000 full-time employees in 14 countries | What was counted Uses generative AI at work without formal approval | Result Over half of the 28 percent who use generative AI at work |
Why employees use AI tools the company has not approved
Employees report using familiar tools and looking for ways to get their work done.
Microsoft UK's October 2025 release reports familiarity with consumer tools and the lack of a work-approved option as reasons for using them. Software AG's September 2024 white paper also reports a preference for choosing one's own tools, alongside employees saying that the business does not provide what they need.
Employees in the KPMG and Melbourne study reported using free public AI tools as well as tools provided by their employer. The way they get a tool does not establish whether its use is approved. The report also records employees who did not know whether their organisation had guidance on generative AI.
The 2024 Work Trend Index from Microsoft and LinkedIn and Software AG's 2024 white paper both report gaps in company training, a reason to check what employees need to learn.
Does banning AI tools work?
These surveys do not establish whether bans reduce unapproved AI use.
In the KPMG and Melbourne study, uploads of sensitive company information or copyrighted material to public AI tools were most commonly reported by employees who said their organisation had banned generative AI. The study cannot show that the bans caused the uploads. Its authors read the pattern as a sign that outright bans may be ineffective and a policy alone does not guarantee compliance.
In Software AG's survey of 6,000 knowledge workers aged 18 and over in the US, UK and Germany, conducted from 13 to 25 September 2024, half of those using personal or non-company tools said they would keep using them even if their company banned them. Closing everything down prevents people from discovering what might help.
How should your company respond?
Find out which work is being done and why employees choose those tools.
Adapt the KPMG and Melbourne questions for an anonymous survey. How often do you intentionally use AI tools in your work? Which do you use: free public tools, tools the company provides or tools you pay for yourself? Is there guidance on generative AI, a ban, no policy, or are you unsure? How often have you used AI at work without knowing whether it is allowed? Follow up by asking which tasks people use those tools for and why they choose them.
If employees are unsure about permission, check that the rules answer their task and that people can find and understand them. If personal tools are being used, ask whether an approved option exists and why it is not being chosen. A missing capability calls for a different response from a preference for a familiar tool. Ask what people need to learn rather than treating the absence of a course as proof of a particular skills gap.
Write the rules as decisions an employee can apply to the document in front of them: which data may go into which tools and how a new tool gets approved. If the approved options do not support the work, investigate an alternative for that task. AI training for employees should use tasks from their jobs to practise checking AI's output and applying the company's rules. In AI training, participants practise in the tools the company has approved and learn where a person must decide.
Sources
- Rise in Shadow AI tools raising security concerns for UK. Microsoft UK, 13 October 2025.
- Trust, attitudes and use of artificial intelligence: A global study 2025. KPMG and The University of Melbourne (Gillespie, Lockey, Ward, Macdade and Hassed), 2025.
- Chasing Shadows: Understanding and Managing Shadow AI, white paper 2024. Software AG, 2024.
- AI at work is here. Now comes the hard part (2024 Work Trend Index). Microsoft and LinkedIn, 8 May 2024.
- More than Half of Generative AI Adopters Use Unapproved Tools at Work. Salesforce, 15 November 2023.
Facts last checked on 27 October 2026.