Canadian businesses struggle with accuracy and workflow in AI adoption push
A push to integrate AI into Canadian workplaces is hitting roadblocks like inaccurate outputs and increased managerial workloads.
TORONTO · Canadian businesses are struggling to integrate artificial intelligence into their daily operations, facing problems with accuracy, workflow disruption and the need for extensive employee retraining. The difficulties were highlighted at a recent Toronto summit for manufacturers, where participants shared stories of delayed projects and unreliable AI outputs. This rocky adoption comes as the federal government pushes for widespread AI use to boost Canada's lagging productivity, with a goal to increase business adoption from 12 per cent to 60 per cent by 2034. Official data shows planned AI use has already soared, but the path to effective implementation is proving complex.
“There’s still a lot of hallucination going on and it’s taking more human reinforcement than we would have liked,” said Connor McCormack, a technical sales specialist at Ontario-based machinery supplier Wittmann Group. He was referring to a chatbot project delayed by accuracy and data safety issues. Other business leaders at the Centennial College event echoed similar frustrations. Deepan Dhanasekaran, an engineering manager with Scarborough-based NMC Dynaplas, said the ChatGPT Enterprise tool keeps producing incorrect data, adding that while it is a good tool, getting it right is a work in progress.
Despite these hurdles, AI adoption rates are climbing rapidly. A Chartered Professional Accountants Canada report from October 2026 found AI use in Canadian workplaces increased from 5 per cent in 2025 to 43 per cent in 2026. Statistics Canada data shows a similar trend, with planned use of AI tools among businesses rising from 10.6 per cent in late 2024 to 25.2 per cent in late 2026. The Carney government's AI for All strategy aims to harness this growth to reverse a decades-long productivity slump, with TD economists estimating faster adoption could boost annual productivity by a full percentage point by 2031.
READ MORE: Loblaw partners with Meta's Muse AI to shop for groceries on customers' behalf
READ MORE: AI tracks individual ovenbirds by their signature songs
Integrating the technology, however, has not been smooth. Wilson Cross, CEO of HR company Borderless AI, described a divided landscape. “It’s a tale of two cities,” he said. “You have these leaders who are very stuck in the old ways and are deploying an old school playbook in a brand new environment. That’s not going to work.” The rollout requires resolving accuracy issues, reorganizing workflows and repeatedly training employees, all while building trust in the new tools. These efforts demand significant time and money, with companies scrutinizing whether the investment translates to real productivity gains.
An emerging challenge is the proliferation of low-quality, AI-generated content. Katie Berg, vice-president of marketing at Vancouver software firm Klue, said she spends much of her time managing what she calls “AI slop grenades.” She explained that people easily produce long, smart-sounding documents using tools like Claude, shifting the burden of critical review to the reader. “You still have to be able to judge what is a good idea versus what just sounds like one,” Berg said. She worries companies will produce bland, generic content for clients and that standing out requires more than just faster tools.
The accuracy problem extends into major corporations, adding layers of work for managers. A senior employee in TD Bank's Toronto operations department said junior staff keen on using generative AI tools like Microsoft Copilot often produce inaccurate content. The manager, who was not authorized to speak publicly, said this forces supervisors to spend extra time fact-checking. In a statement, TD chief AI scientist Maksims Volkovs said the bank expects employees to “validate AI-assisted work before using it,” and that managers review content as they would any other work. TD also uses its internal AI developer, Layer 6, to create custom tools for different departments. For Canada, the consequences of a weak or chaotic AI adoption are significant, as the country continues to trail its G7 peers in productivity. The goal of becoming a global AI leader hinges on businesses handling these early growing pains successfully. As Berg from Klue noted, using AI has compelled her employees to think more about their own skills and how tools can complement rather than replace them, a mindset shift that may be essential for the technology to deliver on its promised gains.
With files from The Globe and Mail