A majority of American workers are running the numbers on their own relevance, and the math is not working in their favor. Fifty-eight percent of U.S. workers say they fear becoming obsolete because their skills are falling behind, according to a survey of 15,357 U.S. adults conducted by The Harris Poll for ETS, an education testing organization, in August and September 2025. The follow-up finding is sharper: 68% say upskilling costs are too difficult to cover, and more than half say getting employer support for training is itself a barrier. The survey's term for this cluster of fears is FOBO: fear of becoming obsolete.

Amit Sevak, CEO of ETS, addressed the structural nature of the problem directly in a statement accompanying the findings:

"America's workers are ready to adapt, but too many are being asked to do it without the time, resources or roadmap they need. That cannot be the worker's burden alone."

The fear is not distributed evenly across the labor market. It reaches 74% among technology workers and 73% among those in financial services, the two sectors where AI adoption and automation are reshaping roles fastest. Bureau of Labor Statistics (BLS) data on occupational change consistently flags both sectors among those with the highest projected skill-disruption rates through 2030, which gives the ETS findings a durable context beyond a single survey cycle.

What the numbers actually say

Four data points from the ETS-Harris Poll survey define the shape of the problem. 58% of workers fear skills-driven obsolescence. 68% say upskilling costs are difficult to cover. 63% say finding time to learn while meeting existing workloads is the binding constraint. And more than half report that accessing employer support in the first place is hard. These are not independent concerns; they compound each other. A worker who can't afford a course, can't get reimbursed, and can't carve out study time is effectively locked out of the upskilling market regardless of motivation.

That motivation is not in question. Close to 9 in 10 workers in the survey agreed that employers should actively encourage staff to set credential goals throughout their careers. Workers are signaling readiness to commit to structured learning; what they are pushing back against is being expected to finance and schedule it entirely on their own.

The quality of training that is available also matters. The Conference Board has noted that when employers do run AI training programs, the content typically covers the basics: prompting and general AI literacy. Advanced capabilities, such as working with AI agents or evaluating AI output quality, are rarely included. The practical implication is that workers completing standard employer-run AI training may find those credentials underweight the skills that live job postings now specify.

For employees: what to do when your employer will not fund the upskilling

If you are among the 58% who sense the ground shifting under your current role, the instinct to immediately sign up for the most prominent course on the market is understandable and usually counterproductive. A more durable first step is to identify where the actual skill gap sits, not where general anxiety suggests it might be.

Three practical steps worth taking this quarter:

• Audit your role against live job postings, not against career anxiety. Pull 10 to 15 recent listings for the job title one level above yours. Skills appearing in 7 or more out of 10 postings represent a genuine market gap. Skills appearing in 2 out of 10 are signals worth noting but not worth paying for yet.
• Make the ask in writing before paying out of pocket. More than half of workers say employer support is hard to obtain, but that figure also means a meaningful share do receive some form of help. A written request that names a specific course, states the dollar cost, and connects the learning outcome to a business result is considerably harder for a manager to decline than a general request for development support.
• Negotiate for protected time, not only budget. 63% of workers identify time as the primary barrier. If an employer declines to fund a course, asking for four hours per week of calendared learning time costs the organization nothing in direct spend and is frequently easier for a line manager to approve than a budget line.

If the answer to all three approaches is a firm refusal, that response is itself useful data. It tells you where your role sits in the organization's development priorities, and it is a reasonable factor in evaluating whether other employers offer better terms. Checking salary benchmarks and verified employee reviews before making a move gives you a view of what you are heading toward, not just what you are leaving behind.

For HR and employers: your upskilling mandate is running into a cost wall

Workforce strategies built on the expectation that employees will take independent ownership of their development are running into a documented affordability problem. Two-thirds of the workforce say upskilling costs are out of reach. Nearly two-thirds say time is a binding constraint. Over half say that when they do try to access employer support, that support is not there. A learning-and-development approach that relies on employee initiative without addressing those three barriers is not a program; it is a policy gap with a good-sounding name.

The downstream effects extend beyond retention. The Society for Human Resource Management (SHRM) released findings indicating that employees at organizations offering upskilling opportunities were significantly more likely to report being engaged at work. Engagement risk and retention risk run together; the ETS data shows the conditions for both deteriorating at scale. WorldatWork's annual Salary Budget Survey also tracks learning-and-development allocations as a share of total compensation investment, and organizations that deprioritize that line item tend to see it reflected in voluntary separation rates within 12 to 18 months.

Three structural shifts worth making

Move the training curriculum past AI literacy basics. The Conference Board's observation that most employer-run AI training stops at prompting and general awareness is a diagnostic that applies to most corporate learning catalogs. Foundational AI literacy is now a baseline expectation, not a differentiator. Programs that do not include a track for integrating AI into existing workflows, working alongside AI agents, and critically evaluating AI outputs are, in practice, training workers for a version of the job that will not exist in 18 months.

Budget protected learning time as explicitly as you budget dollars. A learning-and-development line item without ring-fenced hours attached to it functions as an accounting entry, not a program. The ETS data confirms that time is the constraint workers cite most consistently. The operational fix is calendared learning blocks, locked against operational demand, signed off by line managers as part of workload planning, not as an afterthought to it. The Department of Labor (DOL) has also highlighted workforce training access as a policy priority under the Workforce Innovation and Opportunity Act framework, signaling that federal attention to employer-side training obligations is unlikely to diminish.

Replace development conversations with funded credential roadmaps. Close to 9 in 10 workers said employers should encourage credential goal-setting throughout their careers. That figure is an operational mandate. Moving from open-ended development check-ins to concrete, employee-owned credential plans with employer-funded milestones converts a retention risk into a retention asset. SHRM's research indicates this connection between investment and engagement is measurable, and turnover replacement costs tracked by SHRM consistently run between 50% and 200% of annual salary for skilled roles, making the business case for funded upskilling straightforward to model.

The shared conclusion

FOBO is not a generalized anxiety about technology. The ETS data shows workers are not asking to be insulated from AI or from role change. They are asking for three specific inputs before being expected to rebuild their professional capabilities: a funded budget line, protected hours on the calendar, and a manager who has actually reviewed the development plan. Those are reasonable conditions, and they are measurable ones.

Organizations that treat those three inputs as a shared investment rather than an individual burden will retain the people they need as AI continues to reshape roles. Organizations that keep issuing upskilling directives without the underlying support infrastructure will find that the ETS number keeps climbing, and that the retention and recruitment costs attached to it follow accordingly. For HR decision-makers benchmarking their current programs, the starting question is a practical one: of the 63% of your workforce that says time is the barrier, how many have a learning block on the calendar this week?

Key takeaways

• 58% of U.S. workers fear skills-driven obsolescence; that figure rises to 74% in technology and 73% in financial services.
• 68% say they cannot cover upskilling costs; 63% cannot find the time; more than half cannot access employer support.
• The Conference Board finds most employer AI training stops at general literacy, leaving advanced workflow skills unaddressed.
• SHRM research links employer-offered upskilling to significantly higher employee engagement rates.
• For employees: audit real gaps against live postings, put support requests in writing, and negotiate for protected learning time alongside budget.
• For HR: advance training programs past AI basics, lock learning hours into workload plans, and back credential goals with funded milestones tied to measurable business outcomes.