Canadian workers in heavy industries are embracing artificial intelligence (AI) faster than their employers are providing the training and support needed to use it effectively, according to new research by Punchcard Systems.

The survey found that 49 per cent of employees are using AI on the job but 68 per cent say they have received no training or other formal support from management.

“We found pretty explicitly that workers have gotten ahead of their employers, and we wanted to understand how accurate that assumption was with real data,” Sam Jenkins, managing partner and co-founder at Punchcard Systems told TechNX.

“We were really surprised because we thought perhaps there would be more formalized infrastructure within businesses and turns out it’s much more ad hoc than we had assumed.”

The Edmonton-based software and technology consultancy published the report, The State of AI at Work in Canada, based on a survey of 1,100 working Canadians conducted by Angus Reid Group between Sept. 2 and 11.

Workers still wary of AI output 

While the survey found good adoption by workers in heavy industry, the time savings of 6.1 hours per week was mitigated by the finding that 94 per cent of workers had to recheck all of AI’s output.

“It’s being used quite a bit but the people who are using it are also having to rework their work a lot more," Jenkins said.

"Where that adoption gap comes into play is there may not be trust in the systems. There may not be the support by employers. There may have been experimentation, and that experimentation has resulted in some degree of output but that output was potentially proven incorrect or inaccurate, and so that further reduces the trust in the systems.”

The study showed that 24 per cent found errors in output from AI, with 18 per cent those errors were minor or serious. However, three per cent found those errors could cause a serious safety or financial concern.

The productivity gains weren't universal, however, with 35 per cent of users saying AI had saved them no time.

“So if we only measure the time spent generating output, I think we potentially overstate the gain,” Jenkins said.

While the productivity gains remain unclear, AI is having an impact on workers, with 44 per cent of respondents saying it has had a real effect on their jobs.

Training key to effective AI adoption

Some organizations have trained staff to use AI, but there appears to be a gap among blue-collar workers, according to Jenkins.

“We need to be thinking about what training programs look like at all parts of the organization to make sure that learning is available to absolutely everybody. It’s not a white-collar opportunity. All parts of the Canadian economy deserve to be aware of and leverage these tools to make a more productive workforce.”

That education should be geared specifically to each industry or role.

“Generic, how-to-prompt a chatbot doesn’t help a buyer comparing supplier quotes or a site supervisor reading a spec. I think leadership within organizations need to look at real-world tasks and train people on that,” Jenkins said.

The training should also be integrated into everyday processes.

“I think we should also be potentially using existing meetings, whether or not it’s toolbox talks or safety meetings or standups. These are opportunities for briefings to be happening, and we need to be showing successful moments in terms of productivity and AI to all levels of the organization.”

When thinking about how to integrate AI into everyday tasks, organizations should not simply attach it to an already flawed process.

“Putting AI on top of a bad process is only going to make the bad process faster,” he said.

“In order for AI to be most effective, we actually have to redesign the work to be able to leverage the tool. This means we have to map our workflows a little bit more effectively, and really talk to the people who do the work to find the actual steps where AI would fit, and then the step after it where the human checks.”

As a starting point, firms should deploy AI on “low-risk but high-volume tasks,” Jenkins said.

“We heard that in the survey: lookups, specs, paperwork. It’s close to three-quarters of the respondents said that’s where the value is. If we can start with the high-volume tasks, and in particular the ones that are low-risk, that way we can potentially avoid some of the trust and safety incidences that could happen if we start with the high-risk, high-volume tasks as well.”