Why some workers are embracing AI while others won’t use it

Gallup polling shows AI adoption is rising in U.S. workplaces, with roughly 3 in 10 employees using AI frequently and about two-thirds of AI users reporting a positive productivity impact. But skepticism remains high: about half of employees either do not use AI, and among non-users with access, 46% prefer to work the old way while roughly 4 in 10 cite ethical, privacy, or usefulness concerns. Workers’ job-loss anxiety is also rising, with 18% saying their job is likely to be eliminated by new technology within five years, up from 15% previously.

Analysis

The market is still pricing AI as a productivity uplift story, but the more important second-order effect is internal labor reallocation: the first wave of adoption disproportionately benefits managers, knowledge workers, and firms with high task standardization, while exposing low-trust, highly regulated, or judgment-heavy workflows to resistance and implementation drag. That means the near-term alpha is less in “AI adopters” broadly and more in software, workflow, and compliance vendors that sit between users and the model layer, where adoption is sticky and measurable.

The biggest underappreciated risk is not broad job loss over the next quarter; it is liability leakage over the next 6-18 months as casual users rely on tools they don’t fully understand. Hallucinations, privacy leakage, and bad prompting create a new cost center in legal, healthcare, and finance: audit, governance, and indemnification. That should support cybersecurity, data-loss prevention, and model-governance names, while pressuring point-solution software vendors whose ROI depends on self-serve adoption inside skeptical enterprises.

A more contrarian read is that the current anxiety may actually delay full automation monetization and extend the life of human-in-the-loop workflows. If workers resist embedded AI, enterprise buyers will pay for “assistive AI” rather than replacement AI, which slows headcount displacement but improves software attach rates. Over 12-24 months, the winners are likely to be platforms that can prove reduction in cycle time without increasing error rates; the losers are vendors pushing generic copilots into service-heavy verticals where trust is low and productivity gains are easiest to dismiss.

For positioning, the cleanest expression is to own the picks-and-shovels of controlled AI adoption rather than the broad beta of generative software. The near-term catalyst is enterprise budget season, where governance spend can be justified even if productivity budgets are deferred. If job-loss fears rise further, that can paradoxically help compliance and cybersecurity spend before it hurts labor-intensive end markets.

Key Decisions for Investors

• Long PANW / CRWD on a 3-6 month horizon: use any broad AI enthusiasm pullback to accumulate; risk/reward favors the governance and data-protection layer as enterprise adoption becomes more regulated and error-prone.
• Long MSFT vs short a basket of high-valuation AI application names over 6-12 months: MSFT monetizes distribution and workflow control, while standalone copilots face slower adoption and higher churn in skeptical end markets.
• Initiate a pair trade: long VEEV or OKTA / short a service-heavy software or BPO proxy over 6-9 months, targeting the divergence between trusted workflow automation and labor-substitution names exposed to resistance.
• Buy 6-12 month out-of-the-money calls on CRWD or PANW into any volatility spike tied to AI data-leak headlines; those events should improve budget urgency and compress procurement cycles.
• Avoid chasing pure-play generative AI beneficiaries after momentum bursts; instead, rotate into compliance, security, and enterprise platform names on dips where adoption fear is creating better entry points.