AI training budgets buy hours. The research on older workers says timing is the variable

A doctoral thesis summarized on phys.org argues that for experienced employees who are already burned out, more digital training is not reliably an improvement, and that poorly timed, overly complex or under-supported training becomes an additional job demand that erodes a worker's sense of being able to keep doing the job ^[3][4]. That is an awkward finding for the people writing next year's budgets, because the survey evidence in the same piece shows spending is going up while satisfaction with training is not ^[2][14].

Start with the demand signal. Statistics Canada found that in March, 36% of Canadian workers said they had used generative AI tools as part of their main job or business in the past year, about half said they were familiar with how the tools are used, and 93% were aware of them ^[6][7]. A separate survey cited in the piece put the share of Canadian workers who want or need to upskill to use generative AI effectively at 83% ^[20]. The Canadian Federation of Independent Business found 78% of Canadian businesses planned to maintain or increase training spending in 2026 ^[2]. Against that, a TD Bank survey found only 37% of Canadian workers said their employer had provided adequate training, meaning 63% did not ^[14][18]. Money is moving; the perceived gap is roughly 46 percentage points wide between wanting help and calling the help adequate, across two different surveys ^[20][14][19].

The exposed cohort is the one hardest to replace. Statistics Canada data cited in the article show the share of workers aged 55 and older within the average organization went from 9.3% in 2001 to 18.8% in 2022, an increase of 9.5 percentage points, or roughly 2.0 times the 2001 level ^[8][17]. Since 2000 the total number of Canadian workers aged 55 and over has grown by 184%, faster than any other age group ^[9]. The author's argument for why this matters is not sentimental: early departures take institutional knowledge, mentorship and hard-won expertise with them ^[10].

The underlying work is two studies. The first was a systematic review of 121 articles that identified 14 gaps in what is known about technology's effect on aging workers ^[11]. The second surveyed 361 participants, built on job demands-resources theory, which treats job demands and available resources as the two forces determining employee well-being ^[12]. According to the author, burnout weakened workers' sense of their own work ability, which in turn pushed retirement intentions earlier, and technological training moderated that relationship: how and when training was delivered mattered as much as whether it happened ^[5].

Two mechanisms make the direction of that moderation plausible rather than surprising. Technostress, a term coined by psychologist Craig Brod for the anxiety and fatigue of adapting to new technology, grows when training adds to demands instead of reducing them ^[13]. And research on training transfer holds that whether employees apply new skills depends on their motivation and capacity once they are back at their desks, both of which are scarce in people running on empty ^[21]. Burnout also raises the odds of considering leaving a job outright, and for workers near retirement that decision is exit from the workforce, not a move to a competitor ^[15][16].

Operationally, the implication is that hours purchased is the wrong unit. The tractable variables are sequencing, complexity and post-course support, and whether the recipient has the slack to absorb any of it ^[4][21].