The 79% shadow AI figure doing the rounds in Canadian boardrooms was collected in May 2025, and the exposure has only widened since.
Shadow AI in Canada is usually introduced with one statistic: 79% of Canadian office workers use AI at work and only 25% use enterprise-grade tools their employer provided. The number is real. It is also older than most people quoting it realise. IBM published it on 3 September 2025, from Censuswide fieldwork run between 23 and 30 May 2025. Anything you decide from it is a decision made on a snapshot taken before the current wave of desktop assistants and agent features shipped.
TL;DR
- IBM published the Canadian shadow AI figures on 3 September 2025, with fieldwork run 23 to 30 May 2025 by Censuswide across 4,000 office workers in four countries.
- 79% of Canadian office workers use AI at work, and only 25% rely on enterprise-grade tools.
- 33% mix personal and employer tools, and 21% use personal apps only.
- IBM put the added cost of shadow AI at roughly CA$308,000 per data breach.
- Only 29% believed their employer uses AI to its full potential.
What the IBM shadow AI study found
The IBM release is blunt about the split. Four in five Canadian office workers use AI at work. One in four uses a tool their employer sanctioned and secured. A third run a mix of personal and company tools, and 21% use personal apps exclusively.
That last group is the one that matters. A third of staff mixing tools is a governance issue. A fifth of staff working entirely outside the sanctioned stack is a data issue with a price attached, which IBM put at about CA$308,000 in added cost per data breach.
Why the age of the number matters
May 2025 is several product generations ago. Since that fieldwork, assistants moved from browser tabs to desktop applications with inbox and file access, and agent features arrived inside tools companies already licence. The share of workers using unapproved AI has not plausibly fallen.
Newer Canadian data points the same way. The Bank of Canada reported in August 2026 that more than two thirds of business leaders personally use AI weekly while only 8% of firms use it significantly in core operations. Individual use keeps running ahead of the company. That is the definition of shadow AI, measured from the top of the org chart instead of the middle.
Use the IBM figure as a floor, not a current reading. Then go and measure your own.
What should leaders do next
- Pull your own number this month. Ask every team which AI tools they used last week, anonymously, with an explicit promise that nobody gets in trouble for answering.
- Publish an approved list the same week you ask. A survey with no alternative on offer teaches people to stop answering honestly.
- Check whether staff can install desktop software without approval. That single setting decides how much of your shadow AI has file access.
- Name the data that never goes into an unapproved tool. Client records, payroll, anything under a confidentiality clause. Keep the list to one screen so people remember it.
FAQ
What is shadow AI?
Employees using AI tools their employer has not approved or secured, usually personal accounts on consumer apps, with company information going into them.
Is the 79% figure still accurate for Canada?
It is the best published Canadian figure but it reflects May 2025 conditions. Treat it as a minimum and measure your own organisation.
What is the fastest way to reduce shadow AI?
Give people a sanctioned tool that is at least as good as the one they picked. Policy without a usable alternative moves the behaviour, not the risk.
Sources
- IBM Canada, Shadow AI Use Surges as Canadian Workers Outpace Employers in AI Adoption, 3 September 2025. canada.newsroom.ibm.com
- Bank of Canada, Canadian businesses use of AI, what the evidence shows, August 2026. bankofcanada.ca
- Statistics Canada, Canadian Survey on Business Conditions, third quarter 2026, 31 August 2026. statcan.gc.ca
Related reading
- AI Transparency in Canada Needs a Practical Company Record
- AI Tool Fatigue and the Business AI Switchboard
- AI Implementation in Canada Needs Repeatable Workflows
Disclosure
The author has no relevant financial, advisory, or board relationships with any party named in this brief.
Written by the AI Magazine Canada team, reviewed by the AI Magazine Canada editorial team.
More in the same thread on our AI for business in Canada reference.
