How the Bank of Canada Measures AI Exposure

The Bank of Canada’s 2026 occupational exposure analysis ranks Canadian occupations by how exposed they are to AI-driven disruption, based on how much of each job’s task content overlaps with what current AI tools can already do: drafting, summarizing, calculating, scheduling, and pattern-based analysis.

The Occupations Most and Least Exposed

Office and administrative work sits at the top of the exposure list: scheduling, data entry, document drafting, first-pass writing. Skilled trades and health care sit at the bottom. That split isn’t about which work is more skilled: it’s about which tasks are physical, unpredictable, and dependent on in-person judgment. AI can draft a report or calculate a quote. It can’t yet read a joint that’s out of alignment or make a physical repair on-site.

Only 19.2% of Canadian businesses report using AI in any form (Statistics Canada, Q2 2026), and that adoption is heavily concentrated in finance, insurance, and corporate services, not construction, fabrication, or the trades. That concentration matters: it means the disruption showing up in career-anxiety surveys is real, but it’s also narrower than the headlines suggest. It’s landing hard on a specific slice of the economy, not spreading evenly across every industry at once.

Cross-Checking Against the WEF Data

This lines up with the World Economic Forum’s Future of Jobs Report 2025, which we’ve referenced before: 86% of employers expect AI to transform their business by 2030, and analytical thinking is rated the single most sought-after skill. The same WEF report projects 170 million new roles created globally against 92 million displaced, a net gain of 78 million, but a genuinely disruptive reshuffling in between. That reshuffling is exactly what shows up as anxiety in the survey data: net job growth doesn’t feel reassuring to someone whose specific role is one of the 92 million being displaced.

Read together, the WEF data explains why change is coming, and the Bank of Canada data shows specifically where it’s landing, and skilled trades sit outside the highest-risk zone in both data sets.

What About Robots, Specifically?

It’s worth separating two different questions: AI replacing office tasks, and robots replacing physical labour. The second gets less attention, and the data on it is arguably more reassuring. Canada ranks 13th globally in robot density, with 241 robots per 10,000 manufacturing workers, well behind leaders like South Korea (International Federation of Robotics, World Robotics Report 2025). Robot installations in Canada actually fell 12% in 2025, down to 3,800 units, and most of that activity is tied to the automotive sector’s investment cycles rather than a steady replacement of skilled labour across industries.

What This Looks Like By Process

The exposure data holds at the industry level, but it’s worth being specific about what welding work itself pays, since that’s the practical question behind all of this. Government of Canada Job Bank data lists the national wage range for welders at $22.00 to $47.00 per hour, with a median of $30.00 per hour. Provincially, Alberta’s median sits at $38.00 per hour, British Columbia at $35.00 per hour, and Ontario at $28.00 per hour (2023–2024 reference period).

Different welding processes carry different demand profiles too: MIG remains the most common entry point for fabrication and manufacturing work, TIG commands a premium for precision work in stainless and aluminum, Stick stays essential for field and structural repair, and FCAW is valued in heavy structural fabrication. None of these are more or less “AI-exposed” than each other: the exposure gap in this article is about trades versus office work broadly, not one welding process versus another.

What “AI-Proof” Doesn’t Mean

None of this data should be read as “welding will never change.” Digital weld monitoring, better machines, and some automation are already part of the industry, and will keep advancing. Low exposure doesn’t mean no change: it means the core work still depends on a person physically present, making judgment calls a model can’t yet make remotely. Staying current with new tools is still part of the job. It’s just a different kind of change than watching a task get fully automated out of existence.

What Makes a Career Genuinely Safer

If “safe” isn’t a degree and a desk anymore, the data points to a different profile: work that requires physical presence, hands-on judgment, and skills built through repetition rather than instruction: exactly what a skilled trade teaches. Add a demand curve that isn’t slowing down, and the case for redefining “safe” is hard to ignore.

How This Should Change a Career Conversation

In practice, this data is most useful as a question-reframe. Instead of asking “which career sounds impressive or traditional,” the more useful question is “which career depends on tasks a model can already do.” That single filter reorganizes a lot of career advice that hasn’t caught up to 2026 yet, including advice that still treats a university degree as an automatic safer bet than a trade credential.

The Local Picture in B.C., Alberta & Ontario

This split shows up locally too. In B.C., Alberta, and Ontario (the provinces home to our Surrey, Edmonton, and Brampton campuses), skilled trades remain some of the most in-demand, least AI-exposed careers available, regardless of the broader economic headlines.