Is it really AI hollowing out Canadian junior jobs – or is it hybrid work?

The conversation in Canadian boardrooms has reached something close to consensus. Generative AI is compressing the entry point into white-collar careers. The tasks that once kept junior analysts, associates, and coordinators occupied — drafting, data processing, routine research, first-pass summarisation — can now be handled by a language model, and organisations are acting accordingly.

Three in ten HR leaders in the United States and Canada report that their talent acquisition strategy is shifting toward hiring fewer entry-level workers in favour of mid-level employees, with 56% saying AI has reduced the need for senior staff to assign basic research, writing, data analysis, and administrative work to junior associates. Among 15- to 24-year-olds in Canada, youth unemployment reached 14.7% in September 2025, its highest rate since 2010 excluding the pandemic years.

But a major new working paper — published in May 2026 by researchers at the University of Warwick, the London School of Economics, and Oxford's Ellison Institute of Technology — argues that this consensus may be built on a causal error with significant consequences for policy and practice. The study, spanning 243 million new hire records and 407 million job postings across four Anglophone economies between 2017 and 2025, finds that when the effects of AI exposure and working-from-home exposure are properly disentangled, the WFH effect is robust — and the AI effect largely disappears.

The paper, The Broken Ladder: AI, Remote Work, and Early-Career Hiring, by Peter John Lambert and Yannick Schindler, centres on a methodological problem that has affected virtually every study linking AI to junior hiring decline. Both WFH exposure and GenAI exposure, measured at the occupation level, appear to powerfully predict declines in early-career hiring when analysed on their own. A two-standard-deviation increase in either predicts a fall of roughly 4–5 percentage points in the junior share of new hires by 2025.

The problem is that these two exposure measures are almost interchangeable. The WFH index and the GenAI index have a Spearman rank correlation of 0.77 across 683 occupations. Software developers, accountants, lawyers, and management consultants sit near the top of both. Electricians, truck drivers, and construction workers sit near the bottom. When an exposure-based design attributes the junior hiring decline to AI, it may simply be capturing the WFH effect operating on the same set of occupations.

When both variables are estimated jointly, the WFH coefficient remains stable and significant across every specification. The AI coefficient attenuates sharply and often becomes statistically indistinguishable from zero. A confounder explaining only 1–3% of residual variation in both treatment and outcome would be sufficient to eliminate the AI coefficient. Eliminating the WFH coefficient would require a confounder five times more powerful.

Canada entered the post-pandemic period with one of the highest rates of remote work adoption in the world, and that adoption has not fully unwound. According to Statistics Canada's Labour Force Survey, those mostly working from home amounted to 17.4% of employed people in May 2025, down from 18.7% in May 2024 — still many times above the 4% recorded before the pandemic in 2016. In Toronto, remote work rates are substantially higher: Ottawa-Gatineau recorded 34.2% of workers mostly working from home in 2024, well above the national average.

Robert Half's analysis of over 285,000 new Canadian job postings shows that fully in-office positions declined from 71% of postings in Q4 2023 to 61% in Q4 2025. Hybrid arrangements have risen correspondingly. These are the exact conditions the Lambert-Schindler paper identifies as suppressing junior hiring — not because firms are making explicit decisions to disadvantage early-career workers, but because the organisational conditions that made investing in junior talent tractable have changed.

Canada also faces a specific policy moment that makes the paper's findings timely. The federal government's September 2024 directive requiring core public service workers to work on-site at least three days per week — and executives four days — was justified in part by the onboarding, collaboration, and culture-building benefits of in-person time. That rationale is entirely consistent with the Lambert-Schindler mechanism: if those benefits are real, their absence in hybrid-first private-sector organisations may be materially suppressing junior hiring investment.

The paper's model formalises a dynamic that experienced HR leaders will recognise. Firms hire junior workers not solely for their immediate productivity, but as an investment in the experienced professionals they will become. That investment rests on the firm's ability to develop those workers through proximity: informal mentoring, visible performance, feedback loops that allow rapid skill accumulation, and the gradual development of firm-specific knowledge and professional judgment.

Remote and hybrid arrangements raise the cost of all of these. Supervision requires more deliberate effort. Informal feedback mechanisms break down. The incidental learning that happens by working beside a senior colleague — watching how they handle a difficult client, manage up, navigate ambiguity — does not transfer through a screen with equivalent fidelity. Research cited in the paper found that firm-wide remote work made collaboration networks more siloed; that in-person time delivered the largest productivity gains for the youngest workers; and that proximity to colleagues significantly increases feedback frequency and quality for early-career employees.

As these organisational returns on junior investment fall, the rational firm tilts its hiring upward — toward workers who have already built their capabilities through prior experience and who require less scaffolding to be productive under hybrid conditions. Repeated across thousands of Canadian organisations, that shift produces the junior hiring decline now being attributed to AI.

Research documents that employment for young software developers declined by nearly 20% by July 2025, with similar drops for computer occupations and service clerks. AI is genuinely reshaping what tasks junior workers perform. Demand is shifting toward workers who can supervise, interrogate, and improve AI outputs rather than produce first drafts from scratch — a meaningful change in what employers are looking for.

What the paper challenges is whether AI adoption is the primary reason firms are choosing not to hire junior workers, as opposed to a reason they are choosing to redesign what junior workers do. The D2L survey of Canadian HR leaders found that 64% of respondents expect entry-level hiring to increase over the next 24 months — a finding difficult to reconcile with AI substitution as the sole structural force, but entirely consistent with a WFH-driven friction that organisations can learn to manage.

Data from Challenger, Gray & Christmas found that AI was responsible for 54,836 announced layoff plans in the United States in 2025. But Gartner's own survey of 321 customer service and support leaders found that only 20% had actually reduced staffing because of AI. The gap between stated rationale and actual mechanism is where the Lambert-Schindler paper does its most important work.

The practical implications are more tractable than the AI-displacement narrative suggests. Audit whether your organisation's junior hiring decline tracks most closely with the functions that adopted hybrid and remote work earliest — or with those deploying AI tools most aggressively. The answer is likely more informative than any AI benchmarking exercise.

Review whether your onboarding and mentoring infrastructure has been rebuilt to compensate for the proximity losses of hybrid work. Young workers in 2025 have fewer opportunities for the unplanned, incidental learning that occurs through physical co-presence with senior colleagues — the kind that survey data shows 25% of workers are already worried about losing. Has your organisation designed explicit alternatives?

Consider whether your hybrid policy treats early-career employees as a separate category requiring different structural support — or whether it applies the same framework to a partner and a first-year analyst alike. The paper's model suggests these are fundamentally different employment relationships from the firm's perspective, and managing them identically may be costing more than it saves.

Canada's youth unemployment challenge, its demographic pressures, and its need to build homegrown talent pipelines for AI-era industries all converge on the same imperative: the career ladder must remain accessible at the bottom. If the Lambert-Schindler paper is right, keeping it accessible is less about AI policy and more about management practice. That is a challenge Canadian HR is well placed to meet.