If your AI workforce communication is built around the assumption that employees will be neutral or modestly concerned about AI's effect on their long-term job security, the assumption is off by a substantial margin. 64% of US adults expect AI to lead to fewer jobs over the next 20 years. Only 5% expect more jobs. The remaining 30% are divided between "not much difference" (14%) and "not sure" (16%). Among AI experts surveyed by Pew, the distribution is less extreme but still leans negative: 39% predict fewer jobs, 19% more jobs, 33% not much difference. The public is more pessimistic, but both groups expect AI to reduce overall employment. The workforce sentiment landscape is more negatively skewed than most organisational AI communication strategies engage with.

Views on whether AI will create or eliminate jobs, public vs experts, 2025:

The pace dimension reinforces the level dimension. Experts forecast generative AI will assist 8% of US work hours in 2027, rising to 18% in 2030. The top 25% of expert predictions reaches over 30% by 2030; the top 10% exceeds 40%. The public expects 10% by 2030, substantially lower than the expert median.

Generative AI work-hours assist intensity, public vs expert forecast, 2025–2030:

The two patterns combined produce the structural data observation: experts and the public both expect AI to reduce jobs over 20 years, but experts expect faster AI adoption (more disruption sooner) while the public expects slower adoption (less immediate disruption). The gap in pace is as large as the gap in level severity. Workforce communication needs to absorb both dimensions.

The occupational specificity is informative.

Views on AI-driven job loss by occupation, public vs experts:

On specific occupations, both experts and the public broadly agree that cashiers (US adults 73%, experts 73% expecting fewer jobs), journalists (67% / 60%), and software engineers (59% / 60%) are at high risk for AI-driven job loss. The divergences show where the two groups disagree:

  • AI experts see greater risk for truck drivers (50% vs 48% public) and lawyers (38% vs 23%).
  • The US public sees greater risk for teachers (45% vs 35%), medical doctors (29% vs 27%), and mental health therapists (43% vs 31%).

The public is generally more likely to anticipate job loss across categories than experts, but the categories where the public is most concerned (teachers, medical professionals, mental health therapists) are precisely the human-service categories where experts are less convinced of AI's substitution capability. The public expects more displacement; experts expect displacement to land in different occupations.

The data observation: the public is more pessimistic about AI job loss in general but specifically more pessimistic about AI replacing roles that depend on social, emotional, and ethical judgement. The expert view is closer to "AI will displace specific technical and operational roles; human-judgement roles are harder to displace." The two framings differ on which roles are at risk, not just how many.

[CHART fig_918_2026: global expectations of AI creating vs eliminating jobs by country, 2025; chart not yet built]

The international picture adds context. Country-level expectations about AI creating versus eliminating jobs vary substantially. Nigeria, Japan, Mexico, UAE, South Korea, and India all expect AI to create more jobs than it eliminates (shares above 60%). The US and Canada sit at the opposite end, with 67% and 68% of respondents expecting AI to eliminate jobs and disrupt industries. The US-centred Pew finding of 64% expecting fewer jobs is part of a broader North American pessimism that does not generalise globally.

Three structural implications follow for organisations setting AI workforce strategy in 2026.

The first implication: US workforce AI communication operates in a more negatively-skewed sentiment environment than international workforce communication. The same AI deployment message that lands neutrally or positively in India, Mexico, or the UAE may land as confirmation of feared job loss in the US. Multinational AI deployment plans need country-specific communication that engages with the local sentiment baseline rather than applying a single global template.

The second implication: the occupational specificity of public concern is operationally actionable. Organisations deploying AI in the high-public-concern occupations (teachers, medical doctors, mental health therapists) need communication strategy that engages explicitly with the substitution question rather than relying on general augmentation framing. The public is sceptical that AI will be additive in these roles; communication that asserts augmentation without engaging with the substitution concern will be received as evasion.

The third implication: the pace gap (experts faster, public slower) creates a specific organisational risk. Organisations that deploy AI quickly to capture productivity benefits are operating on the expert pace timeline. The public is anchored on the slower timeline. When AI deployment outpaces public expectations, the surprise factor compounds the negative sentiment. The workforce-management implication: pacing AI deployment to match public expectations (or communicating accelerated deployment proactively) reduces the surprise compounding effect. Organisations that deploy at expert pace without explicit communication will trigger more workforce friction than organisations that deploy at the same pace with deliberate communication.

The data observation: US workforce sentiment about AI job impact is structurally more negative than expert sentiment by a substantial margin, the negative skew applies most strongly to human-service occupations, and the international pattern shows the US is at the more-pessimistic end of a global distribution rather than the global norm. The 2026-2028 trajectory of public sentiment depends on whether actual AI deployment validates or contradicts the public expectation of broad job loss. If deployment produces visible displacement in the occupations the public is most worried about, sentiment will harden. If deployment is more selective and creates new role categories alongside displaced ones, sentiment may moderate. The trajectory is not yet determined.

For organisations setting US workforce AI strategy in 2026, the data anchor is that 64% of employees expect fewer jobs in 20 years. The communication strategy that engages with this expectation directly, acknowledging the concern, addressing the specific occupational fears, and providing visible evidence of how the organisation's AI deployment is producing role evolution rather than just role elimination, will be more effective than communication that operates from the expert framing of AI as augmentation. The data is clear; the implications for communication strategy are direct.

Sources

  • Primary: Stanford AI Index 2026, Chapter 9 (Public Opinion) 9.2 — hai.stanford.edu/ai-index/2026
  • Public-expert job views: McClain et al. (2025) — Pew Research Center US survey of 5,410 adults and 1,013 AI experts, 2024
  • Recent public job view: Kennedy et al. (2025) — Pew Research Center US survey of 5,023 adults, June 2025
  • Work-hours forecast: Forecasting Research Institute LEAP 2026 — Longitudinal Expert AI Panel work-hours assist forecasts
  • Cross-country comparison: Ipsos AI Monitor, 2025 — multi-country AI sentiment survey