Study: Over 2 in 3 Canadian Heavy Industry Workers Received No Formal AI Support From Their Employers

Punchcard’s The State of AI at Work in Canada Report found that construction workers are the least likely to receive formal AI support from their employer

EDMONTON, Alberta--(BUSINESS WIRE)--AI is making its way onto factory floors, construction sites, and farms across Canada, but most heavy industry workers have been given no formal guidance on using the technology, according to new research from Punchcard Systems, an Edmonton-based digital innovation consultancy.

Nearly half (49%) of Canadian heavy industry workers already use AI for work, yet over two in three (68%) have received no formal AI support or training from their employers. That figure rises to 87% of construction workers, the least supported sector in the study.

This is according to Punchcard's The State of AI at Work in Canada Report, developed in partnership with Angus Reid. The report surveyed 1,100 working Canadians across a wide variety of industries to understand how AI is changing their day-to-day work and the tasks where they trust AI the most or the least.

"The people closest to the work already know where AI helps and where it creates risk," said Sam Jenkins, Managing Partner at Punchcard Systems. "When employers give them proper guidance and a real say in how it's used, leaders get a much clearer picture of what AI can do for their organizations. The most meaningful AI gains come from rethinking how the work gets done, rather than bolting an AI tool onto outdated workflows and processes."

Is AI saving Canadian workers time, or creating more work?

It's doing both. Workers who save time with AI gain back hours each week, but nearly everyone who uses it still has to fix what it produces, with many catching AI errors before they cause an issue at work.

• Over half (52%) of Canadian workers overall use AI for work. However, the adoption rate varies by industry (49% in heavy industry vs. 57% in the service sector).
  • Public admin and government workers are the most likely to use AI for work (76%), while automotive workers are the least likely to adopt the technology (36%).
• Canadian workers who save time with AI save an average of 6.1 hours per week.
• Nearly all (94%) Canadian workers rework AI-generated output before they can use it.
• About one in six Canadian workers (18%) say AI-generated output has caused a problem at work, while 24% say AI errors were caught before they caused a problem.

Are Canadian employers training workers to use AI?

For most workers, the answer is no. Construction workers are the least supported sector, with 87% receiving no formal AI support, followed by agriculture (77%), manufacturing (67%), and transportation and logistics (65%).

The lack of support persists even as AI has impacted many Canadian workers’ jobs.

• More than four in ten (44%) Canadian workers overall say AI has impacted their role, rising to 54% for transportation and logistics workers, 52% of professional services workers, 45% for manufacturing workers, and 45% for oil and gas workers.
• For many, the biggest change is a new layer of review work added to their day. About one in five (19%) say reviewing or fact-checking AI-generated work is a regular part of their job.
• One in ten (10%) workers say their employer tracks or monitors their AI use, such as token usage, number of prompts, or activity inside AI tools.

Only 6% of workers who use AI say they've been given formal responsibility for AI governance, policy, or safe use at their organization.

Workers whose employers provide formal AI support are three times as likely to hold that responsibility (9% vs. 3% of those whose employer has provided none), suggesting that the gap in governance is closely tied to the broader gap in support.

Which tasks do Canadian workers trust AI with?

Mostly routine ones. Workers are comfortable handing paperwork and lookups to AI, but many draw the line at decisions that carry physical, financial, or safety risk.

• Heavy industry workers trust AI most for looking up routine procedures, product specs, or how-to information (77%) and writing up reports, logs, or job paperwork (71%).
• Nearly half (45%) of heavy industry workers don't trust AI at all with decisions that affect worker safety.
• Construction workers are among the most willing to trust AI on high-stakes tasks, as long as they can review what it produces. Nearly two in three construction workers trust it to read technical drawings (64%) and interpret Canadian codes and standards (63%).
• While 36% of construction workers trust AI on decisions that affect worker safety as long as they review it, none say they would trust it completely with those safety decisions.

These findings suggest the next stage of AI adoption in Canada will depend on how well organizations keep people involved in the decisions AI touches, and whether their tools and rules are built for the work their teams actually do. Punchcard works with organizations to modernize outdated systems and build custom AI tools, with clear rules on where a person needs to make the call.

Access Punchcard’s The State of AI at Work in Canada Report here: www.punchcard.io/reports/state-of-ai-canada-heavy-industries/

About Punchcard
Punchcard Systems is a Canadian digital innovation consultancy that helps organizations turn complex problems into purposeful, human-centred software solutions. We partner with over 150 clients across Canada and the U.S. to push the boundaries of what thoughtful technology can do for business, community, and society. Learn more at punchcard.io.

Survey Methodology
These are the findings of a survey conducted by Angus Reid Group on behalf of Punchcard from September 2-11, 2026 among a sample of n=1,100 working adults living in Canada who are employed full-time, part-time, or are self-employed/contractors, and are members of the Angus Reid Forum. The survey included a national sample of Canadian workers, with additional targeting of employees working in heavy industries, including manufacturing, construction, natural resources, electric power/utilities, automotive, and agriculture. For comparison purposes only, a probability sample of this size would carry a margin of error of ±3.0 percentage points, 19 times out of 20.