AI Will Barely Touch Your Payroll This Year, Then the Math Changes
The honest read on AI and Canadian jobs just arrived from the central bank, and it is calmer than either the boosters or the doom crowd would like. Canadian firms told the Bank of Canada that AI will do very little to their headcount over the next 12 months. Push the clock out to three years and the same firms expect a modest net negative effect on employment. Not a cliff. A drift. That drift is the part worth planning around now, because three years is one budget cycle plus a hiring freeze, and it arrives faster than most workforce plans assume.
TL;DR
- The Bank of Canada’s June 2026 staff paper used special AI questions in the December 2025 Business Leaders’ Pulse survey.
- Business leaders use AI personally all the time. Adoption for actual production is still limited.
- Firms expect AI to lift capital spending over the next 12 months, and a touch more over three years.
- Employment impact is close to zero this year. Over three years, firms expect a modest net negative.
- The signal for operators is timing. The labour effect is slow, real, and worth building into your plan before it shows up.
What happened
On June 2, 2026, the Bank of Canada published staff analytical paper 2026-22, “Survey Evidence on Firm AI Adoption and its Implications,” written by Chanya Chawla and Crystal Arnburg. It draws on special questions the Bank added to its December 2025 Business Leaders’ Pulse, the survey it runs to take the temperature of Canadian firms.
The headline finding is a split screen. Personal use of AI among business leaders is widespread. Using AI to actually produce goods and deliver services is not. Most firms are still running AI in the margins of the business rather than the core of it.
On money, the leaders lean positive. On balance they expect AI to push their capital spending up over the next 12 months, and slightly more over three years. That tracks with what you would expect from a technology people are buying into before they have rebuilt anything around it.
On jobs, the read is restrained. Firms anticipate limited employment effects over the next year. Over three years they expect a modest net negative impact on headcount. The Bank’s own summary calls AI adoption among Canadian firms early stage, with the bigger economic effects expected to show up later.
Why it matters in Canada
Canada has been waiting for AI to fix a productivity problem that predates the technology. Output per hour worked here sits near the bottom of the G7. The hope was that AI would close some of that gap fast. The Pulse data says firms themselves do not expect fast.
Pair this with the firm-level work Statistics Canada published in April 2026. That study found AI adopters look 16.8% more productive than non-adopters on paper. Control for how productive those firms already were, and the premium drops to 10.2%. Add controls for the other capabilities adopters tend to have, things like R&D, cloud computing, data analytics, and staff training, and the AI effect falls to 5.1% and stops being statistically significant. Read plainly, the productivity edge belongs to the whole operating model, not the AI license sitting on top of it.
So the central bank survey and the StatCan microdata point the same way. AI is real, the payoff is slow, and it shows up only inside firms that have done the surrounding work. For a Canadian operator that is a more useful map than another forecast about millions of jobs vanishing next quarter.
Business impact
Here is what the numbers mean for the people who actually sign off on hiring and capital.
The capex signal is the near-term one. Your peers expect to spend more on AI-related capital this year. That is budget pressure whether or not you have a plan, because vendors and boards read the same surveys you do. The risk is approving spend on tools while the operating model stays put, which is exactly the pattern that produced flat returns across the past two years.
The employment signal is the slow one. A modest net negative over three years does not mean layoffs announced on a Friday. It looks like a role that does not get backfilled when someone leaves. A team that absorbs 20% more volume without a new hire. A junior tier that gets thinner because the first draft now comes from a model. Spread that across a mid-market company and you get a payroll that grows slower than the business, which is the understated version of a jobs effect and the one your competitors will use to widen margins.
Picture a 200-person Canadian firm. Nothing dramatic happens this year. Over three years, normal attrition runs 15% to 20%. If AI lets you backfill four in five of those exits instead of every one, you have changed your cost base without a single difficult meeting. That is the math the Pulse data is hinting at, and it favours the operators who see it coming.
What leaders should do next
- Separate your AI capex from your AI plan. If you are approving spend this year, attach each dollar to a workflow you intend to rebuild, not a license you intend to hand out.
- Run a three-year headcount model with and without AI assumptions. Use normal attrition as the lever. Decide on purpose which exits you backfill, rather than letting it happen by accident.
- Fund the complements, not only the tool. The StatCan evidence is blunt. Data analytics, cloud, R&D, and training are what convert AI into productivity. Fund those alongside any AI purchase.
- Protect your junior pipeline. The slow jobs effect tends to land hardest on entry-level roles, which is also where your future managers come from. Decide what you are doing about that before the gap opens.
- Watch the next Business Leaders’ Pulse for movement. If the three-year employment expectation gets more negative, the timeline you are planning against just got shorter.
The skeptic’s view
A fair-minded executive can push back here, and should. Survey expectations are not outcomes. Firms are guessing about three years out, and business leaders have a long record of over-predicting how fast technology will change their cost base. The Bank itself frames these as early-stage expectations, not forecasts. If AI follows the path of past general purpose technologies, the labour effect could be slower and gentler than even this restrained number suggests, with new roles appearing to offset the ones that thin out.
There is also a case that the modest net negative is mostly composition, not destruction. Firms grow, work shifts, and some roles fade while others appear. Treating a three-year survey expectation as a layoff plan would be a misread, and cutting staff now on the strength of it would be a genuine mistake.
The counter is simply that planning for a slow change is cheap and planning for none is not. You lose very little by modelling the headcount math early. You lose real ground if the drift arrives and your competitors modelled it and you did not.
What to watch
- The next Business Leaders’ Pulse release and any updated AI questions, for whether the three-year employment expectation shifts.
- The Bank of Canada’s summer and fall 2026 monetary policy commentary, where AI and productivity keep surfacing.
- Statistics Canada’s next business conditions release, for whether production-stage AI use climbs past the low double digits.
- Federal execution on the AI for All strategy’s adoption targets, which assume a much faster jump than firms currently expect.
Closing analysis
The most valuable thing in this data is its tempo. AI is not going to rescue your productivity numbers this year, and it is not going to gut your payroll either. It is going to move slowly, reward the firms that built the surrounding capability, and show up in employment as a payroll that grows slower than the business rather than a round of cuts. The operators who win the next three years are the ones treating that as a planning input today, while it is still cheap to get ahead of.
Sources
- Bank of Canada. “Survey Evidence on Firm AI Adoption and its Implications.” Staff Analytical Paper 2026-22. June 2026. https://www.bankofcanada.ca/2026/06/staff-analytical-paper-2026-22/
- Statistics Canada. “Artificial intelligence adoption and productivity in Canadian firms.” Economic and Social Reports. April 22, 2026. https://www150.statcan.gc.ca/n1/pub/36-28-0001/2026004/article/00002-eng.htm
- Statistics Canada. “Analysis on expected use of artificial intelligence by businesses in Canada, third quarter of 2025.” 2025. https://www150.statcan.gc.ca/n1/pub/11-621-m/11-621-m2025011-eng.htm
- OECD. “Macroeconomic productivity gains from artificial intelligence in G7 economies.” OECD Artificial Intelligence Papers, No. 41. 2025. https://www.oecd.org/
Disclosure
The AI Magazine Canada blog has no relevant financial, advisory, or board relationships with any party named in this column.
