Canadian firms struggle to turn AI trials into ROI

BDO Canada has published research showing that many Canadian businesses are still struggling to turn artificial intelligence trials into measurable returns. The report found that 46% of business leaders are experimenting with AI without meaningful ROI.

Only 18% of respondents said they were actively embedding AI into workflows and operations, while 27% said AI would have minimal impact on their organisation over the next four years. The findings are based on a survey of 520 Canadian business leaders from the Angus Reid Forum.

The figures point to a divide between interest in AI and the harder task of putting it to work in day-to-day operations. BDO argues that the challenge has moved beyond pilot projects and now centres on governance, workforce readiness, and the ability to link AI spending to clear business outcomes.

The issue comes as more businesses prepare for the spread of agentic AI systems, which are designed to handle multi-step tasks and interact across software platforms. Gartner has forecast that by 2028, one-third of enterprise software applications will include agentic AI, up from less than 1% in 2024.

That shift is likely to increase pressure on companies to show they can supervise AI tools in the same way they already demonstrate cyber controls and compliance processes. In BDO's view, governance is becoming a buying criterion for clients and regulators rather than a narrow internal safeguard.

From pilots to operations

The report describes what it sees as a second-stage problem in Canadian AI adoption. After an initial wave of experimentation, many organisations now face questions about ownership, data quality, risk management, and whether projects are producing business value.

Those issues become more important as AI moves beyond stand-alone chat tools into systems that support decisions and automate parts of workflows. In that setting, companies need to determine what problems they are trying to solve, who is accountable for outcomes, and how results will be measured.

Bill Syrros, National AI Leader, BDO Canada, set out that concern in the report's findings.

"The next gap will not be between organizations using AI and not using AI. It will be between those redesigning work around AI and those funding disconnected pilots. As AI moves beyond the chat phase and into workflows, leaders will need to be clear on what problems they are solving, who is accountable for the outcome, what data can be used, whether that data is reliable enough to support the work, and how value will be measured. Most executives do not yet know what that looks like in practice, and the distance between those who do and those who do not is widening," he said.

His comments reflect a broader market concern that enthusiasm for AI has run ahead of operational discipline. Businesses have spent heavily on experimentation, but many are still working out how to integrate tools into established processes without adding risk or confusion.

Governance pressure

BDO argues that organisations will need formal governance frameworks if they want to scale AI use across the business. It pointed to standards such as ISO 42001 as an example of the kind of structure companies may increasingly need as customers, auditors, and regulators ask for evidence of AI controls.

That reflects a wider change in corporate technology oversight. Cybersecurity reviews, supplier assessments, and compliance checks are already standard in procurement and risk management, and AI controls may now be moving into the same territory.

The firm also argues that AI adoption should be treated as an operating-model change rather than a simple technology deployment. In practice, that means changes to employee training, workflow design, accountability structures, and measurement systems, rather than just new software purchases.

Jeff Chapman said the opportunity for Canadian companies will depend on whether they can move beyond tests and into disciplined implementation.

"This is one of the most significant shifts our clients will navigate this decade. Canada's AI opportunity will depend not only on adoption, but on adoption that is practical, measurable and trusted. Our goal is to help Canadian organizations move from experimentation to responsible scale with the same human-led strategy we have built inside our own firm. There is real opportunity for those that build the right foundations, and we are investing to help our clients seize it," he said.

Internal lessons

BDO says its view is informed partly by its own use of AI inside the firm. It describes an internal approach of testing, learning, and scaling AI within operations, alongside investment in workforce enablement, workflow integration, and responsible AI practices.

That experience has shaped the advice it gives clients, particularly around setting risk thresholds for use cases and avoiding internal disputes that can stall projects before they reach production. The report argues that governed execution is more likely to help companies move from isolated trials to repeatable use across the business.

For Canadian employers, the message is that AI uptake alone is no longer the main benchmark. The more difficult test is whether businesses can make the technology useful, accountable, and visible in financial results as it becomes embedded in software and routine work.

One of the clearest survey findings was that more than a quarter of business leaders still expect AI to have minimal impact on their organisation over the next four years.