Companies that train their workforce in artificial intelligence risk losing the investment when those employees accept offers elsewhere. Firms that refuse to train them risk losing the employees first. The dilemma, framed openly by the source publication as a live editorial question rather than a settled answer, has sharpened as AI moved from optional to operational across nearly every corporate function.

The question facing talent leaders is no longer whether to teach AI skills, but how to teach them without subsidising competitors. Neither choice is free. Training broadly accepts a measurable turnover risk. Training narrowly accepts a capability gap and a morale cost among those excluded from the programme.

What does the calculation cost?

Training budgets rarely appear in headline attrition numbers, but the math is familiar to anyone who has managed a learning programme. The recovered benefit shrinks when a significant share of trainees depart before the curriculum pays back. The hidden cost, however, is the cost of doing nothing: teams that cannot use AI competently will eventually need to be rebuilt or replaced at a premium.

The savings do not all leave at once, either. Improved processes, documented workflows and peer-led coaching tend to persist beyond any one resignation. The question is not whether the trained worker leaves, but whether the trained organisation retains what it learned. Companies that document their AI workflows as they build them keep the method even when they lose the practitioner.

What does the workforce tell employers?

The underlying logic is uncomfortable in two directions. Employees who lack AI skills may struggle to remain effective in their roles, which gives employers a reason to train them. Employees who gain AI skills become more attractive to other employers, which gives employers a reason to hesitate. The article's framing as a question, rather than a recommendation, suggests the editors see no settled answer, and the absence of a recommendation is itself information.

The standard HR response to a skills gap is to recruit. The standard HR response to a retention risk is to compensate. AI training sits awkwardly between the two, behaving like both and like neither.

What should HR leaders actually do?

A defensible approach treats AI training as a two-tier proposition. The first tier covers broad, transferable skills — prompt writing, output review, basic model literacy — that any professional can carry to a new employer. The second tier covers proprietary workflows, internal data handling and tool-specific integrations that have value only inside the firm. Train generously on the first. Train selectively on the second. The line between them is rarely tidy, but it exists, and writing it down forces a conversation that most firms have avoided.

The sceptical question remains. If the goal is to retain trained people, training alone is rarely the lever. Compensation, career path and managerial quality tend to move retention numbers more reliably than any curriculum. AI training can sharpen a worker's contribution, but it does not, by itself, give that worker a reason to stay.

Takeaway: employers who treat AI training as a retention tool are likely to be disappointed; employers who treat it as an operating requirement may still lose staff, but they will lose a more capable workforce in the process.

Forward-looking: the firms that answer this question well will not be those that train the most, but those that decide, on paper, which skills are theirs to keep and which are theirs to lend.