AI hiring hype meets a cooler reality as US job gains stay modest
US employers added 178,000 jobs in March, according to the latest figures from the Bureau of Labor Statistics – another month of steady but unspectacular growth. The numbers barely differed from February, even as corporate leaders continued to promote artificial intelligence as a catalyst for productivity, innovation, and new jobs.
The contrast between the rhetoric and the data is becoming harder to ignore. While executives talk about AI-fueled expansion, recent labor and workplace surveys suggest a far messier transition, marked by weak tech hiring, shrinking entry-level opportunities, and rising frustration among rank‑and‑file employees.
—
Job growth comes from old-economy sectors, not AI darlings
Most of March’s job creation came from traditional, non-tech sectors. Healthcare led the way, adding around 76,000 positions. Construction followed with an estimated 26,000 new jobs, while transportation and warehousing expanded payrolls by roughly 21,000. Social assistance roles also grew, underscoring the continued demand for in‑person, service-oriented work.
By contrast, tech-adjacent categories showed little momentum. Employment in areas such as computing infrastructure and web search portals was essentially flat. More telling, computer systems design and related services – a core slice of the tech economy – shed about 13,000 jobs in March.
That pattern undercuts public claims that tech hiring is firmly back on track. High-profile investors and executives have argued that fears of AI-triggered job losses are exaggerated, pointing to rising job postings at technology companies. Yet the official data shows that, at least for now, much of the tangible hiring is happening outside the AI and software heartland.
—
Job openings vs. actual hiring: a widening gap
One explanation for the disconnect is the growing gap between open roles and completed hires. Companies may advertise positions to signal growth, attract investors, or build future candidate pipelines, but that does not guarantee that they will fill those jobs at scale or at speed.
The March numbers highlight this tension. Sectors loudly touting AI-led transformation did not translate those narratives into broad-based employment gains. Instead, industries with more immediate, physical needs – hospitals, clinics, building sites, warehouses, and social services – continued to do most of the hiring.
This divergence raises a sharper question: Is AI currently creating enough new work to offset the jobs it displaces or slows down? Early evidence suggests the answer depends heavily on occupation, education level, and where a worker sits in the corporate hierarchy.
—
Entry-level roles feel the squeeze
The pressure is particularly acute at the bottom rung of the ladder. A recent Goldman Sachs analysis, reported in industry coverage, estimates that AI has eliminated roughly 16,000 jobs per month over the past year. In parallel, a 2025 SignalFire study found that hiring of new graduates has fallen by about 50% relative to pre‑pandemic levels.
SignalFire captured the shift bluntly: the once wide‑open door into tech for fresh graduates is now “barely cracked.” According to the report, a combination of smaller funding rounds, leaner operating models, contraction in graduate programs, and increased reliance on AI tools has tightened the market for junior talent.
For young professionals, this means fewer entry points into high-paying, high-growth careers – especially in software, data, and digital services. When AI can draft code, generate reports, or handle parts of customer support, employers may feel less urgency to hire and train people with minimal experience.
—
Displacement into lower-quality work
Goldman Sachs also warned about what happens to workers who are nudged out of tech or higher-skill roles. Many do not exit the labor force entirely; instead, they shift into more routine, less specialized jobs. These positions often pay less, offer fewer advancement opportunities, and make less use of the skills workers previously accumulated.
Over time, this kind of downward occupational mobility can erode the value of prior education and experience. A software tester who becomes a generic operations clerk or a content specialist who moves into basic administrative work may see both wages and skill relevance decline. The report suggests that such transitions can weigh on workers’ earnings and career trajectories for years.
This pattern complicates rosy narratives that displaced employees will simply “reskill” into better, AI-augmented work. While that is possible for some, the emerging data implies that many are instead landing in roles with less bargaining power and lower long-term returns.
—
Executives praise AI; employees live its downsides
Despite these frictions, enthusiasm at the top remains strong. Survey data cited by Harvard Business Review indicates that about 80% of executives are already using AI at least once a week, and nearly three-quarters say they are seeing positive returns from early deployments in their organizations.
On the ground, the story looks different. Mercer found that 43% of workers reported that their jobs have become more frustrating, not less, in the wake of AI adoption. Another survey from Workday estimated that for every 10 hours of supposed efficiency gains attributed to AI, nearly four hours are consumed fixing errors in AI-generated outputs.
This gap between boardroom optimism and frontline experience is one of the clearest fault lines in the current AI rollout. Leaders tend to see dashboards, summary metrics, and pilot success stories; employees confront buggy tools, unclear expectations, and added responsibilities for catching machine mistakes.
—
The rise of “workslop” and the cost of rework
Harvard Business Review also highlighted a new phenomenon dubbed “workslop” – content that appears polished and professional but lacks depth, coherence, or substance. AI systems can generate such outputs quickly and confidently, which makes them particularly insidious.
Researchers found that 41% of workers had encountered this kind of low-quality yet convincing work. Each instance triggered nearly two hours of additional review, editing, or complete rewrites. Rather than eliminating grunt work, AI often reshapes it into a different kind of invisible labor: checking, validating, and cleaning up after automated systems.
Workday’s data reinforces that point. Only 14% of respondents said they consistently achieve net-positive outcomes from using AI at work, meaning that in most organizations, the gains are uneven, situational, and sometimes fully offset by rework and oversight.
—
The “AI tax” on everyday operations
For teams tasked with operational stability and high accuracy, the burden can be especially heavy. While senior leaders deploy AI for strategy documents, brainstorming, and synthesizing information – contexts where occasional errors are tolerable and creativity is prized – operational staff often have to integrate AI into workflows where mistakes carry real costs.
Brian Solis of ServiceNow has described this drag as an “AI tax”: more checking, more rework, more anxiety. Customer support agents must verify AI-crafted responses. Finance teams must scrutinize AI-assisted analyses for subtle numerical or logical errors. HR staff must ensure that AI-generated job descriptions and performance reviews are fair, factual, and legally compliant.
The result is a paradox. Tools introduced under the banner of productivity can, at least in the short term, slow teams down, increase cognitive load, and heighten the sense of risk. Until systems become more reliable and processes are redesigned, many workers experience AI less as an assistant and more as an extra layer of work.
—
Why the disconnect persists
Several structural factors help explain why AI’s promised benefits have not fully materialized in the labor statistics:
1. Learning and integration costs
Organizations need time to redesign workflows, train staff, and build guardrails around AI tools. During this transition, efficiency often drops before it improves.
2. Mismatch between tools and tasks
AI performs well on pattern recognition and language generation, but many jobs involve interpersonal nuance, physical coordination, or domain-specific judgment that current systems struggle to replicate.
3. Risk and compliance constraints
In regulated industries such as healthcare, finance, and law, the tolerance for AI-generated errors is low. That limits how far and how fast automation can go, and requires extensive human oversight.
4. Uneven access and skills
Not all workers are equally prepared to leverage AI effectively. Those with strong digital literacy and domain expertise can gain leverage; others may feel sidelined or overwhelmed.
These dynamics mean that AI can be simultaneously overhyped and genuinely transformative, depending on which part of the labor market one examines and over what time horizon.
—
What this means for workers now
For employees and job seekers, particularly those early in their careers, the current environment calls for a more strategic approach:
– Prioritize complementarity, not competition
Roles that pair human strengths – empathy, negotiation, complex problem-solving – with AI capabilities are more resilient than tasks that can be almost entirely automated.
– Invest in transferable skills
Data literacy, critical thinking, and the ability to interpret and challenge AI outputs are increasingly valuable across functions, from marketing to operations to product development.
– Seek organizations with mature AI practices
Employers that define clear AI policies, provide training, and measure actual productivity impacts are more likely to deliver net benefits, rather than shifting the AI tax onto individual workers.
– Stay alert to “hidden” rework
Workers should be encouraged to document how much time is spent correcting AI errors. That data can inform better tool selection, workload planning, and expectations at the leadership level.
By proactively engaging with AI rather than ignoring or fearing it, workers can carve out niches where their skills are amplified rather than eroded by automation.
—
The policy challenge: safety nets for a faster cycle
Even AI developers acknowledge that the employment landscape is changing in ways that policy is not fully prepared to handle. OpenAI has floated broad policy concepts such as expanded healthcare coverage, stronger support for retirement savings, and a renewed industrial agenda designed to cushion workers during technological transitions.
The company emphasized that these ideas are preliminary, primarily intended to spark debate. It also issued a warning: if public policy fails to keep pace with the rapid deployment of AI, the institutions and safety nets meant to absorb shocks could lag dangerously behind.
That concern echoes historical patterns. Previous waves of automation reshaped entire sectors – from manufacturing to retail – often faster than retraining systems and social protections could adapt. The question now is whether governments, employers, and educational institutions can move quickly enough to avoid repeating the most painful parts of that history.
—
A more realistic AI narrative
Taken together, the March jobs data and recent surveys point to a more nuanced story than either utopian or dystopian extremes. AI is not yet delivering a broad hiring boom in the United States, especially in tech and entry-level roles. It is, however, already reshaping who gets hired, which tasks are valued, and how work feels on a day‑to‑day basis.
Executives may be right that AI will ultimately unlock significant productivity gains and new kinds of jobs. But the early stages of adoption are revealing friction, displacement, and a growing divide between those who benefit from AI’s promise and those who bear its costs.
Bridging that gap will require more than bold claims of “AI-led growth.” It will demand transparent measurement of outcomes, honest accounting of rework and displacement, targeted upskilling, and policies that recognize the human side of technological change – not just the efficiency gains that show up on quarterly slides.