Most Canadian companies are paying for AI and cannot show any Canadian AI ROI. 93% of Canadian business leaders say their company uses or pilots AI. 2% report a measurable return on it. That gap belongs on your operating dashboard, and most boards are not asking about it.
Key points
- KPMG Canada surveyed 753 business leaders in late 2025. 93% use AI in some form. Only 2% report measurable ROI.
- 31% have pushed AI into core operations. The rest are still piloting or partially deployed.
- 57% of leaders say their biggest problem is figuring out how to capture value, up from 40% a year ago.
- Productivity wins are showing up. Revenue and cost wins mostly are not.
- The bottleneck is workflow design and governance, not the tools.
What happened
Canadian AI ROI is rare: 93% of leaders say their firm uses AI, and only 2% report a measurable return.
| Signal | Share of leaders | What it means |
|---|---|---|
| Use AI in some form | 93% | AI is common; value is not |
| Measurable ROI on gen AI | 2% | Almost nobody can show a return |
| Fully integrated in core operations | 31% | Most firms are still partial or pilot |
| Hardest problem is capturing value | 57% (was 40%) | Workflow and governance, not more tools |
KPMG Canada’s Generative AI Business Adoption Survey, published November 2025, polled 753 business leaders between August and September. AI use jumped from 61% to 93% in a single year. Only 31% of those leaders said their company has fully integrated AI into core operations. Another 32% have partial deployment. 20% are still in early pilots. 17% are stuck at experimentation.
2% reported measurable ROI. Most of those were companies with over $1 billion in revenue. Among the few seeing a return, 31% could not even quantify it.
The same survey found 57% of leaders now say their hardest problem is understanding how to capture value from AI, compared to 40% the year before. Fewer than four in 10 have a written plan for how AI will create value.
Statistics Canada’s Canadian Survey on Business Conditions for Q2 2025 paints a quieter version of the same picture. Only 12.2% of Canadian businesses report using AI to produce goods or deliver services. That is double the year prior, but still small. Of those that did adopt, 40% built new workflows around the tool and 39% trained staff on it. Almost 90% reported zero change in headcount.
In May, Bank of Canada Governor Tiff Macklem warned that the productivity boost from AI will roll out slowly and that Canadian firms taking a tepid approach are at risk.
Why Canadian AI ROI matters
Canada has a productivity problem, and AI was supposed to be part of the answer. Statistics Canada data shows AI-using firms outproducing non-adopters by a wide margin once value capture starts. The CFIB estimates Canadian SMEs save just over an hour a day from generative AI use, and that reinvesting half of that time could lift GDP by $12.8 billion.
That dividend is not arriving for most Canadian businesses. The KPMG numbers say almost everyone is buying the tools. Almost no one is converting them to dollars on the income statement.
The regulatory picture is part of the context. Bill C-27, which carried the proposed Artificial Intelligence and Data Act, died on the Order Paper when Parliament was prorogued in January 2025. Canada now has no federal AI law. Provinces are moving on their own, and Quebec’s Law 25 and the EU AI Act are setting the practical bar for any Canadian company with cross-border data flows. Boards are running AI programs on PIPEDA, written in 2000, plus a patchwork of provincial rules.
This is the environment your AI investment has to earn its keep in. Buyers are getting smarter. Capital is tighter. American competitors are scaling fast. Your 2% of measurable return is what your board is going to ask about next quarter.
Business impact
You can read the KPMG number two ways. The optimistic read is that 2% is what early ROI looks like in any technology cycle and the rest will follow. The realistic read is that most Canadian companies have spent two years buying tools and licences while the operating model around those tools stayed put. That is the read to bet on, and it is a judgment, not a measured finding.
What this looks like inside a typical mid-market company in Canada:
- Sales reps run prompts in ChatGPT for emails, but the CRM workflow is the same as it was in 2023.
- Finance has a Copilot subscription, but month-end close still takes the same number of days.
- HR drafts job descriptions in an AI tool, but time to hire has not moved.
- Customer support runs an AI chat layer that deflects a few tickets, but staffing models are unchanged.
Activity is up. Output is roughly flat. That is the ROI gap in operational terms.
The KPMG data shows another tell. Among AI adopters, the top spend is on hiring tech talent and buying more tools. Change management and adoption are getting one-third of the attention. That ratio is upside down for the value capture stage Canadian businesses say they are in.
What leaders should do next
- Pick three workflows where you will actually rebuild the process around AI. Skip “use AI in” sales. Rebuild lead routing, qualification, and outbound from the ground up. Same for one finance workflow and one customer-facing workflow.
- Set a measurable target for each. Cycle time reduction, cost per ticket, revenue per rep, days to close. If you cannot pick a number, you do not have a project, you have a license.
- Audit your tools. Most companies are paying for three to five overlapping AI products their staff barely use. Cancel the duplicates.
- Write an AI value plan in one page. KPMG found 62% of Canadian companies do not have one. Yours can.
- Put governance and use policies in writing before staff do it for you. Shadow AI is already inside your company. Quebec Law 25, Ontario Bill 194, and your customer contracts already create liability if you do not.
- Invest in training the way you invest in tools. The CFIB found AI-adopting businesses are 5.4 percentage points more likely to fund employee training. The companies seeing returns are in that group.
The skeptic’s view
There is a fair case for patience. Productivity gains from past general purpose technologies took years to show up in the data. Electricity and computing both moved through long periods of investment without measurable return. The Bank of Canada has acknowledged this directly. Some of the 2% number is a measurement problem, not a value problem. Many wins from AI today show up in employee experience, decision quality, and risk reduction, which traditional ROI frameworks miss.
It is also reasonable to wait if your industry has not produced strong reference cases yet. Spending on AI for the sake of looking modern has burned plenty of Canadian companies in the past 24 months. Doing less, more carefully, is a defensible call for a CFO right now.
The argument for moving anyway is competitive. Canadian firms are already lagging US peers on AI adoption depth. Waiting another two years to redesign a workflow is two more years of US competitors compounding their gains.
What to watch
- The next KPMG Canada AI adoption update, expected in late 2026. Watch if the 2% ROI figure moves.
- Statistics Canada Q3 and Q4 2025 business conditions releases. They will show if AI adoption is broadening past the 12.2% baseline.
- Federal signals on a successor to AIDA. The current Parliament has not tabled a new AI bill, but the file is alive at ISED.
- Ontario’s Bill 194 implementation, which will set the practical AI governance bar for any company doing business in the province.
Closing analysis
The 2% reflects how Canadian businesses are buying AI, more than it reflects AI. Tools without workflow redesign, licence counts without training, pilots without an ROI plan. That is the operating problem. The few Canadian companies capturing value are doing unglamorous work on process and people. The forecast here is that the next 18 months will separate firms that do that work from firms that keep adding to the AI budget line with no payback line beside it. What would prove this wrong is a later KPMG survey showing the 2% rising while workflows stayed unchanged. Finance shows the same split, where AI in finance in Canada rests on 53 large companies.
Why the national number looks so different
Statistics Canada’s second-quarter 2026 business survey puts AI use at 19.2% of businesses, counting only AI used to produce goods or deliver services. That is far below the 93% of leaders who say they use AI in some form, and the gap is a matter of definition: a chatbot on a laptop counts in one survey and not in the other.
The same survey found 44.4% of AI users changed training or staffing practices because of AI. For an owner measuring a return, that is the practical test: count the AI that has changed how work is done, and leave the rest out of the ROI sum.
Frequently asked questions
How many Canadian businesses see a return on AI?
Very few. In KPMG’s survey, 93% of Canadian business leaders said their company uses or pilots AI, but only 2% reported a measurable return, per KPMG Canada.
Why is Canadian AI ROI so low?
Most firms add AI on top of old processes and do not measure a baseline. Without redesigned work and clear metrics, savings are hard to find and harder to prove.
How can a business measure AI ROI?
Record time and cost for a task before AI, run the same task with AI for 30 days, and compare. Count licence, training and review time as costs.
Related reading
- Shadow AI Inside Canadian Companies and What To Do About It
- How To Write a One Page AI Value Plan for Your Business
- What Bill C-27 Dying Means for Canadian AI Compliance in 2026
Disclosure
AI Magazine Canada Blog has no relevant financial, advisory, or board relationships with any party named in this column.
This piece sits inside our reference on AI for business in Canada, which we keep current.
BDO Canada later put a second number on the same problem. See why AI ROI goes missing when nobody redesigns the work.
