1 in 3 Workers Report a Ghost Downsizing As AI Shrinks Their Teams — But Most Bosses Disagree

1 in 3 American workers say their team has gotten smaller in the past three years while their workload stayed the same or grew. Among mid-level employees, who carry the operational weight of most organizations, that figure rises to 33%. In hybrid environments, where AI adoption is most intense, it reaches 36%.

The Technology sector is the sharpest case study in the data. 38% of Tech and Media workers report team compression, the highest rate of any sector. Simultaneously, 28% report that their team has actually grown, making Technology the most volatile workforce environment in the study. It is shrinking and expanding at the same time, restructuring faster than any other industry, and offering workers the least stable ground to stand on.

Research from the Yale School of Management published in May 2026 describes the same pattern: firms are not cutting headcount broadly, but are getting more output from the same workforce, meeting headcount targets by slowing or freezing the replacement of workers who leave [1].

The executives tell a different story, at least officially. 53% of C-suite respondents across all sectors say AI has not affected their organization’s hiring decisions. Among Tech executives specifically, that figure drops to 30%, the lowest of any sector. Tech leaders are the least likely to claim immunity from AI’s workforce effects, while simultaneously leading every sector in active headcount reductions (17%) and hiring freezes (11%).

Healthcare sits at the opposite extreme. 75% of Healthcare executives report no hiring impact, which matches what workers in that sector report: the lowest compression rates and the lowest AI-related anxiety of any industry.

Looking ahead, 24% of all executives expect AI to drive headcount reductions or hiring freezes within three years. In Tech, that figure is 28%. When current reductions, existing freezes, and anticipated adjustments are combined, 56% of Tech executives are already operating or planning to operate with AI-driven workforce changes.

AI is not landing equally across the income spectrum. This survey measures that gap as a reflection of how technology is being experienced right now by people at different income levels within the same labor market.

For workers earning between $100,000 and $175,000, AI functions as an administrative partner. 67% describe their AI exposure as augmentation: the technology makes them faster and better at work, but they are still directing. 60% feel empowered when using AI. 0% report resentment. This cohort has the lowest rate of non-use and the highest rate of productivity excitement at 44%, and not a single respondent in this income band believes more than half their tasks will be automated within five years.

For workers earning under $60,000, the same technology reads as a threat. Only 47% feel empowered, a 13-point gap from high earners. 9% report active resentment, describing AI as undermining their expertise or autonomy. 26% expect AI to automate more than half of their daily tasks within five years.

This aligns with Goldman Sachs Research findings published in April 2026, which found that the negative effects of AI on job creation are falling largely on younger, less experienced workers, the same cohort most represented in lower-income brackets [2].

Education follows the same gradient. Workers with a Master’s degree or higher report 42% being excited about higher productivity and 13% skepticism about AI’s benefits. High school graduates report 26% productivity excitement, and 21% who see no benefit at all. Income and education together define an AI experience that is, at the extremes, barely recognizable as the same phenomenon.

If AI adoption continues to concentrate its benefits at the upper end of the income distribution while concentrating anxiety and displacement risk at the lower end, the workforce impact of this technology will compound existing inequality rather than distribute opportunity across it. The behavioral data already reflects this: lower-income workers are not simply more worried; they are already moving and changing careers at nearly three times the rate of their higher-earning counterparts.

The most statistically significant finding in this study is the relationship between where a worker physically performs their job and how exposed they feel to AI displacement.

Fully on-site workers are the most psychologically secure group in the survey. 40% strongly disagree that their skills will be obsolete within three years. Only 3% are in what the data classifies as acute panic, strongly agreeing that obsolescence is imminent. They report the lowest anxiety when using AI at 17% and the lowest resentment at 4%. For on-site workers, physical presence functions as a cognitive shield.

Remote workers are experiencing a completely different reality. 36% fear their skills will become obsolete within three years. 13% are in acute panic, more than four times the on-site rate. They report the highest anxiety when using AI at 28%, the lowest empowerment of any work arrangement at 41%, and the highest job loss fear at 73%. (p = 0.000000000385, the highest statistical significance finding in the study.)

Remote workers report AI surveillance at 10%, more than three times the on-site rate of 3%. Their outputs are more visible, more measurable, and easier to compare with AI-generated alternatives than those of workers whose value is partly expressed through physical presence. They are not imagining the exposure. They are working inside it.

Hybrid workers represent the clearest outlier in the dataset. At 62% empowerment, they are the most positive group about AI of any work arrangement, higher than both fully remote and fully on-site workers. They report the lowest acute panic at 4% and a job-loss fear rate of 58%, which is closer to on-site workers than to remote ones. Whether the hybrid arrangement represents a durable equilibrium in the AI era or a transitional advantage is a question this data cannot answer. What it can confirm is that the gap between hybrid and remote outcomes is large, consistent, and statistically robust.

Small businesses are managing AI transitions with relative smoothness. Large enterprises are not. That difference is one of the clearest structural findings in the executive survey, and it has direct consequences for the workers inside those organizations.

Among executives at organizations with fewer than 50 employees, 54% report that their workforce has been broadly receptive to AI changes. Adoption is fast, informal, and largely working. 19% of small-business leaders have not yet communicated their AI plans to employees, but at that scale, workers can read the situation through proximity to leadership.

At enterprise scale, the picture changes completely.

Among organizations with 5,000 or more employees, broad workforce receptivity collapses to 27%. 69% of enterprise executives are actively managing employee resistance in some form, and 64% name workforce disruption and morale as top concerns, nearly double the rate for any other company size tier. The same organizational properties that slow AI adoption also slow communication and change management. At scale, the gap between what leaders are planning and what workers understand is too large to bridge informally.

These are the same sectors where worker anxiety is most diffuse and where the absence of clear information leaves the most room for fear to fill the gap. Across all executives, only 23% have deployed a formal AI preparation program with a dedicated budget, and 40% are addressing it informally or on an ad hoc basis. The workers most likely to fear skill obsolescence are the least likely to be in organizations with a structured plan to address it.

Workers and their leaders converge on one question with near-identical results: what actually protects a worker from AI displacement.

Both groups place physical or hands-on work first: 58% of workers and 58% of executives. Both place strong interpersonal and communication skills second: 55% of workers and 51% of executives. The consensus holds across sectors, seniority levels, and company sizes. It is the most stable finding in both surveys, and it runs counter to the prevailing public advice on AI-proofing a career, which tends to emphasize technical upskilling and digital fluency above all else.

The divergence appears further down the list. Executives rank creativity and strategic thinking fourth at 38%, well above the 27% of workers who share that view. Leaders place a significantly higher premium on creative and strategic capacity than the employees who are supposed to be deploying it. That gap may reflect how executives understand value creation differently from how workers experience their daily tasks.

Both groups dismiss seniority. Workers rank it last at 14%; executives rank it last at 16%. In a labor market where institutional knowledge has historically been the most reliable career moat, its near-unanimous dismissal as a protective factor is a finding that deserves to sit without elaboration.

There could be a word for what 30% of American workers are experiencing, and only 18% of their executives will acknowledge.

Ghost Downsizing.

Teams shrink through attrition that goes unfilled, workloads expand to fill the gap, and AI handles what the missing colleague used to do. On paper, employment looks stable, but in practice, the people who remain are doing more with less while their leaders, officially, say nothing has changed.

Because nothing, officially, did.

The people facing the Ghost Downsizing have no announcement to point to, no restructuring plan to reference, no headline that describes what their working day has become.

Every one of these findings leads back to the same individual question: what does this mean for me?

Our AI Job Risk Calculator answers that question at the task level. Rather than categorizing entire professions as safe or at risk, it analyzes the specific skills, activities, and work context of your occupation, along with your seniority, salary, location, and work mode. The result gives you a map of where you stand.

The worker survey polled 665 currently employed US adults via Prolific. Respondents were screened for active employment across full-time, part-time, self-employed, and freelance status, ages 18 to 65. The executive survey polled 354 C-suite leaders, VPs, directors, and business owners with five or more employees, also via Prolific. Both surveys used US-representative sampling with regional segmentation across the Northeast, Midwest, South, and West. Pearson’s chi-square tests were applied across all demographic subgroups, with findings reported at the 95% confidence level. Respondents who failed an embedded attention check were removed before analysis. Demographic data was collected via Prolific prescreeners and matched to survey responses via Participant ID.

1. The Real Job Destruction from AI Is Hitting Before Careers Can Start, Yale.
2. The Jobs AI is Likely to Boost and Those It May Disrupt, Goldman Sachs.