Canadian Workers Say AI Saves Time. Who Gets the Hours?

A task that used to swallow an afternoon can now take a fraction of the time. That doesn't mean the workday ends early. Someone still decides whether the extra capacity goes toward doing the job better, taking on more work or giving people room to breathe.

The Bank of Canada's October 19 survey offers a revealing glimpse of that choice. About 40% of workers surveyed said they use artificial intelligence on the job, mostly generative AI. Among those users, the estimate was striking: they believed they would need about five more hours each week to finish the same work without it.

Nasty Canuck's AI series opener looked at the infrastructure and capabilities Canada needs to put the technology to work. The latest survey brings us to the other end of that chain: what changes for the people actually using it?

Five hours, with an important asterisk

The Bank collected online survey responses from August 5 to 27, with follow-up phone interviews continuing into early September. Workers using AI said they typically used it three or four days a week and spent less than an hour actively using it on those days. They described drafting emails, translating material, checking code, finding information and tackling complicated problems.

The five-hour figure is an estimate supplied by the workers, not time recorded by an employer or an independent productivity experiment. It does not mean every Canadian employee gained five hours, that the working week got shorter or that checking an AI-generated answer takes no time. A draft produced in seconds may still need careful review.

A separate Statistics Canada survey published in July found that 35.9% of workers aged 15 to 69 had used generative AI at work over the preceding 12 months; 41.6% had used at least one AI or automation technology. Those figures should not be lined up with the Bank's 40% as if they were consecutive readings of the same measure. The survey populations, questions and periods differ.

Better work, more work, or some time back

The Bank also asked what AI users did with the time they believed they saved. Improving the quality of their work was the most common answer. Taking on additional tasks followed closely. Roughly a third apiece reported using the time for better work-life balance, learning or training, or more creative work. Respondents could report more than one use; these are not pieces of a pie that must add up to 100%.

All of those can be legitimate gains. A cleaner report, a faster response to a customer or time to learn a new skill may be worth more than an earlier departure. An employer who buys an effective tool will reasonably expect a return, too. The friction comes when the tool speeds up a task but the workload expands so quickly that nobody can tell whether the workplace has actually improved.

There is a difference between resolving a customer's problem faster and simply closing more tickets. One measures a useful result; the other may only measure activity. To know which is happening, an organization has to look beyond the number of drafts produced or buttons clicked.

Canada's productivity dividend is still a question

Statistics Canada's April study of AI adoption and productivity offers a reason to be demanding about results. In its initial comparison, firms adopting AI appeared more productive. After accounting for prior productivity and complementary capabilities such as data analytics, cloud computing, research and staff training, the remaining association was not statistically significant.

One qualification matters: that firm-level analysis draws on earlier technology-adoption data, including surveys from 2019 and 2021. It cannot serve as a direct verdict on the newest generative AI tools or the Bank's 2026 worker responses. What it does show is why existing skills, systems and management practices have to be considered before attributing a productivity advantage to the software alone.

The same caution applies to the five-hour estimate. Multiplying it by every AI user and announcing a windfall for Canadian GDP would be a neat calculation—and a bad one. Time saved on a particular task is not automatically additional output for the economy.

The benefits and the worry are uneven

There is a less cheerful finding in the same Bank of Canada survey. Workers in sectors where AI could replace more tasks expressed concern about job security and were less confident about finding another job. The Bank measured their perceptions, not actual layoffs caused by AI. Those worries still deserve to sit beside the accounts of greater efficiency.

The opportunity to use these tools also varies widely. Statistics Canada's March worker survey found that 65.6% of workers in professional, scientific and technical services had used generative AI at work, compared with 16.3% in accommodation and food services. A national adoption figure conceals different jobs, different tools and different opportunities to acquire experience.

Younger workers face a particular question: how do you learn the judgment behind a task if software now produces its first draft? Statistics Canada's employment research found that employment generally increased across occupations with varying AI exposure through December 2025, while growth for younger workers was weaker. The researchers could not isolate AI's effects from population changes and other economic pressures. Our earlier youth-employment article made the case for better routes into work; this evidence sharpens the question about what newcomers will actually be taught once they get there.

Measure the work, then decide who benefits

A Canadian employer considering AI could start with something more useful than a software-adoption tally. How long did the whole task take before and after, including checking and correcting the result? Did quality improve? Were customers served better? Did staff get time for training, or did the next assignment simply arrive sooner? Those are measurable questions, even if the answers will differ by workplace.

Workers need a say in that discussion because they are often the ones who discover where a tool succeeds, where it invents an answer and where it quietly shifts work to someone else. Employers need clear rules for approved tools, sensitive data and who remains accountable for the finished result. The organization also needs a plan for teaching new colleagues the skills that cannot be learned by accepting every first draft.

There is no single correct way to share a productivity gain. Some workplaces may be able to serve more customers, pay people better, invest in training or ease an exhausting workload. Others will need the extra capacity just to stay competitive. But a claimed improvement ought to lead to an observable outcome, not disappear into another round of targets nobody can explain.

Canada has spent considerable energy debating the infrastructure needed to run artificial intelligence. That investment will be easier to defend if people can point to work that is safer, better or more useful because of it. A report finished before lunch is progress. A workplace that improves without forgetting how to train its next generation is a stronger result.

Editor's note: This is Part 3 of Canada's AI: From Power to Payoff. Part 2, Canada's AI Build Needs a Power Deal Canadians Can See, examines the electricity costs and commitments behind Canada's AI expansion.