What AI is actually doing to jobs.
The data behind the headlines. Canadian firm-level employment impact from Statistics Canada, task-level augmentation vs automation from the Anthropic Economic Index, and global 2030 projections from the WEF Future of Jobs Report. Nothing modeled, nothing fabricated. That includes the parts that complicate the story.
of Canadian firms that adopted AI report total employment stayed the same or grew
Tasks are changing. Jobs mostly aren't.
Same firms, same survey wave (StatsCan CSBC Q2 2025). Left: how much AI reduced tasks previously done by employees. Right: what happened to total employment as a result.
The two charts don't contradict. They explain each other. AI is genuinely reducing tasks for the vast majority of adopters (85%). But task reduction isn't job elimination. Employees whose tasks are reduced get redeployed to other work, and firms hire back what they save elsewhere. That's augmentation at the firm level, and it's the pattern in almost 9 out of 10 AI-adopting Canadian firms.
Augmentation dominates automation across occupations
Anthropic Economic Index — June 2026 report. Task-level analysis of ~1M anonymized Claude conversations mapped to O*NET occupations. Augmentation = AI helps a human; Automation = AI does the task with minimal human input.
57% augmentation vs 43% automation overall. Augmentation dominates in every occupation category where humans do knowledge or judgment work. The occupations closest to 50/50 (Computer & Mathematical, Office & Administrative) are exactly where you'd expect: highly structured tasks that AI can do end-to-end. Even those aren't majority-automation.
Firms are training their existing people, not replacing them
StatsCan Q2 2026 · Changes to training or staffing practices among businesses using AI. Multiple selections allowed per firm.
32% train existing employees. Only 11.4% hire new AI-skilled staff. When a firm adopts AI, the operational response is roughly 3× more likely to be “upskill who's already here” than “go external.” That reframes the AI-and-jobs conversation from “will I be replaced?” to “will I be trained?”, which is a much better question to ask your employer.
AI-skilled workers are being hired right now, at a premium
Real-time snapshot from the Canadian job market. 247,659 active job postings across Canada today; the AI-mentioning subset and its salary premium below.
Roughly 1 in 12 active Canadian job postings mentions AI. Mean salary is 27% above the all-jobs baseline for anything AI-adjacent, and the premium widens sharply for specialized roles. That's live market signal, not a projection. When employers say they're competing for AI-skilled talent, this is what they mean.
Expectations are shifting slightly more cautious
Firms’ own forward-looking view of what AI will do to their total employment. 2024 vs 2025 CSBC waves.
One trend worth flagging honestly: expected decrease ticked up from 9.4% (2024) to 12.2% (2025), while expected increase dropped from 11.1% to 7.3%. The dominant expectation is still “no change” (roughly 70%), but firms are getting slightly more cautious about AI's employment upside. Watch this line in the Q2 2026 release.
Net +78M jobs by 2030, but different jobs
World Economic Forum — Future of Jobs Report 2026. Global aggregate projection to 2030 across ~1,000 surveyed employers representing 14M workers.
Net +78M globally, but the jobs are not the same jobs. Displacement concentrates in clerical, admin, cashier, and data entry roles, which sit in the lower half of the income distribution. Creation concentrates in AI engineering, data science, cybersecurity, and care work, which need either advanced technical training or physical human presence AI can't replicate. The macro number reads optimistic. The transition risk for the 120M workers facing medium-term redundancy is real. Both are true.
What all this actually means
Three defensible reads once you strip out the headline noise.
Three things the data actually says, once you strip out the headline noise:
1. At the firm level, AI is augmenting, not replacing.Nine out of ten Canadian firms adopting AI report the same or more employment. That’s not a projection. It’s a retrospective measurement of what already happened.
2. At the task level, augmentation still dominates.Anthropic’s analysis of a million real Claude conversations shows 57% augmentation, 43% automation. Augmentation dominates in every knowledge-work occupation.
3. Net globally, more roles will be created than destroyed, but the transition is where the risk lives.WEF projects +78M net by 2030. That macro number obscures that clerical, admin, and data-entry roles are genuinely shrinking while AI/data/care roles are growing. The 120M workers whose skill sets are becoming outdated need pathways. That’s the actual policy and training conversation.
The optimism is defensible. The alarmism isn’t. But neither is complacency about the middle 15% who won’t make the transition easily. That’s the tension the data actually shows.
