Reskilling Isn’t the Whole Story

For the past two years, the standard response to anxiety about AI and automation has been simply to tell workers reskill.

Learn new tools. Take a course. Pivot fast enough and you’ll be fine.

It’s an appealing idea, but it’s not the whole story.

Reskilling, which involves learning an entirely new set of skills in order to move into a different role or occupation, is not, on its own, a guaranteed path to a new job. It is often an individual coping mechanism, and one that increasingly asks workers to absorb the risk while offering few assurances about what comes next.

The promise sounds simple. The reality isn’t.

Across governments, companies, and platforms, reskilling has become the default answer to job disruption. Online certificates abound. AI bootcamps promise employability. “AI literacy” is framed as the new baseline skill.

Yet multiple workforce surveys tell a more ambiguous story. In its Global Workforce of the Future research, PwC finds that while many workers are actively upskilling—adding new capabilities on top of existing roles—access to training and its benefits remain uneven. Those who received structured support often report gains in job security or wages, showing that effort alone does not guarantee opportunity.

What’s missing from the conversation is a harder question: reskill into what?

Labor markets don’t reward learning in the abstract. They reward demand.

Much of the optimism around AI relies on headline numbers. The most frequently cited projection—popularized by the World Economic Forum’s Future of Jobs Report 2025—suggests that AI and related technologies could contribute to roughly 170 million new roles globally by the end of the decade, even as around 90 million existing jobs are displaced.

The report is careful, however, to stress that these are net and aggregate figures. They do not indicate where those jobs will appear, how accessible they will be, or whether displaced workers can realistically move into them.

This distinction matters. Job-posting data analyzed by organizations like the OECD and LinkedIn Economic Graph consistently shows that roles requiring AI-related skills are growing faster than average, but remain concentrated in specific sectors, regions, and seniority levels.

For many workers, this does not feel like expansion but compression: fewer mid-level roles, flatter career ladders, and more responsibility pushed onto fewer people.

This tension between optimistic projections and lived experience echoes a broader pattern explored in our Hype vs. Reality analysis of AI and work.

For workers trying to stay employed—not reskill into entirely new professions—the most viable paths often sit around AI systems rather than deep inside them.

These include AI content reviewers who assess outputs for accuracy and reputational risk; model evaluators and red-team analysts who stress-test systems for bias and failure modes; data annotators and AI training specialists who help shape how models learn; AI policy analysts and compliance officers translating regulation into operational rules; product operations managers integrating AI tools into real workflows; and human-in-the-loop supervisors responsible for monitoring and correcting automated decisions.

These roles are increasingly visible in hiring data. Both McKinsey’s research on AI adoption and the OECD’s work on AI-exposed occupations point to growing demand for oversight, governance, and operational roles—often outpacing demand for pure engineering positions.

What these jobs tend to reward is not advanced coding, but judgment, communication, domain expertise, and accountability.

This is where the reskilling narrative starts to fracture.

Workers are encouraged to acquire entirely new skills without being shown how those skills map to durable roles. Certifications are earned, but job descriptions remain vague. Training options are abundant, but pathways to employment are not.

Mid-career workers feel this acutely. Research cited by the OECD and the International Labour Organization shows that older workers face steeper barriers when transitioning into new technical roles—even when they attempt to reskill. Time, savings, and tolerance for risk are finite. Advice that ignores this reality risks sounding less like empowerment and more like abdication.

If AI or automation has affected your work, you’re welcome to share how it’s played out for you.

You can do so in the comments below.
If you’d rather not comment publicly, you can also reach out via the Contact page.

Thanks for taking the time.

There is an uncomfortable reality embedded in much of today’s advice: many of the roles held up as “AI opportunities” involve maintaining, supervising, or correcting systems that have already displaced human work.

For someone who has lost a job to automation, being told to reskill into reviewing AI outputs or managing AI workflows can feel less like a solution and more like capitulation—working for the very systems that made you expendable. That is not an emotional overreaction, but rather a rational response to a loss of agency.

There is also a durability problem. Many of these roles exist precisely because AI systems are still error-prone, poorly integrated, or legally risky. As models improve, some of this work may shrink or disappear. Asking workers to bet their future on positions defined by today’s system limitations is a risky proposition, especially for those without long financial runways.

Both concerns deserve to be taken seriously.

Still, rejecting “AI caretaker” roles does not mean rejecting adaptation altogether. There is a meaningful distinction between serving AI systems and using AI as a tool—and that distinction matters for long-term dignity and stability.

In many professions, the most viable path forward is not reskilling—starting over in a new occupation—but upskilling: learning how to instrument AI within existing roles. That means using AI to compress routine work while retaining responsibility for judgment, accountability, and outcomes.

For experienced writers, editors, analysts, and researchers, this can mean using AI to accelerate early drafts or identify surface patterns, while remaining responsible for structure, accuracy, and narrative coherence. For operational and administrative professionals, it can mean automating coordination and documentation while focusing on handling exceptional cases, stakeholder judgment, and decision-making. In regulated or high-risk environments, it often means treating AI as an assistant rather than an autonomous actor.

This is not reskilling in the sense of starting over. It is upskilling, which means extending existing expertise so that it remains relevant and valuable in AI-mediated workflows.

A narrower, more honest ambition

None of this guarantees stability. No article should pretend otherwise.

But the goal does not have to be becoming “future-proof,” a phrase that increasingly rings hollow. A more realistic aim is to buy time—time to preserve agency, to wait for clearer pathways in the labor market to emerge, and to protect one’s professional identity until the turbulence of AI-driven change stabilizes—and to avoid unnecessary displacement, staying attached to one’s professional identity while the ground continues to shift.

That may mean learning selectively. It may mean resisting some roles while accepting others as transitional. It may mean using AI tactically rather than embracing it ideologically.

For workers navigating this moment, skepticism is not a failure to adapt. It is a survival skill.