No Canadian data supports that, and a chunk of what makes someone a supervisor — signing off on a hazard assessment, deciding whether to stop work — is a duty legislation assigns to a person, not a tool.
Short answer
No evidence points that way. Statistics Canada's own employment analysis found that jobs generally kept growing through the AI adoption wave, with a stated caveat that the cause isn't fully settled. More concretely, several of a field supervisor's core duties under provincial OHS law — investigating a work refusal, reporting a critical injury — are written as duties of a specific person, not a system.
StatCan's analysis of employment through the early period of generative AI adoption (November 2022 to December 2025) found that “employment generally grew regardless of potential occupational exposure to and complementarity with AI,” though growth varied by age and education. StatCan's own caveat matters just as much: “it is unclear whether more recent trends reflect the advent of AI, other economic factors… or a combination of factors.” Read that as evidence against a wave of AI-driven layoffs so far — not proof nothing will change.
A chunk of what a field supervisor does is a statutory duty assigned to a named person. Under Ontario's OHSA, an employer or supervisor investigating a work refusal must do so “immediately, in the presence of the worker,” together with a worker-side representative — a sequence of real-time judgment calls, not a report generated afterward. The fatality and critical-injury rules run on the same logic: a written report within 48 hours follows immediate notification by a person on site, and the incident scene generally isn't to be disturbed except for a narrow, listed set of reasons. It's not an Ontario-only pattern, either: Alberta's OHS Code puts the same person-not-system requirement on hazard assessment, requiring an employer to “involve affected workers in the hazard assessment and in the control or elimination of the hazards identified” — a consultative step that has to happen with the actual crew on the actual site, not against a dataset.
What AI realistically takes off a supervisor's plate is the paperwork around the judgment calls — drafting the incident report, summarizing a toolbox talk, flagging a JHSC recommendation nearing its 21-day response deadline — not the decision itself. That's the same split this hub keeps landing on: AI drafts, a person signs. More on the routine behind it: JHSC minutes and follow-through.
A 30-minute call is enough to map what's safe to automate on your sites and what isn't.
