Workers who want AI training the most are using it the least

AI adoption in the workplace is highest among college graduates and high earners, with usage at 66.3% for workers earning over $200,000 versus 15.9% for those under $50,000. The New York Fed also found that 62% of workers expect AI to raise unemployment, while only about 16% report employer-provided AI training despite 38% saying it matters. The article points to potential labor-market inequities and a need for broader upskilling, but near-term employment impact still appears limited.

Analysis

The market is still pricing AI as a near-term margin lever, but the more important second-order effect is labor stratification: the workers most able to capture productivity gains are also the ones with bargaining power and existing access, while the rest face a widening capability gap. That creates a two-speed adoption curve, where enterprise AI ROI is likely to concentrate in already-high-productivity functions first, then spill into lower-wage roles only if employers fund training rather than just tool licenses. The result is less a broad labor-market shock in the next 6-12 months than a gradual reallocation of surplus toward firms that can operationalize AI fastest.

This is bearish for labor-intensive, low-differentiation service businesses that rely on white-collar throughput but have weak training cultures, because they will either eat higher wage bills to retain trained staff or lose productivity to slower adoption. It is bullish for vendors selling workflow orchestration, compliance, and enterprise deployment layers rather than pure model access, since training and change management become the gating factor. The underappreciated winner is probably not the frontier model provider, but the picks-and-shovels stack that helps firms convert employee intent into measurable output.

The main risk to the bearish labor narrative is timing: the employment effect is likely a 12-24 month story, not a single-quarter earnings catalyst. What can reverse it is a corporate training catch-up cycle or a macro slowdown that forces management teams to treat AI adoption as a cost-out mandate rather than an optional perk. A genuine capex-to-labor substitution wave would show up first in guidance language, then in hiring freezes and reduced backfill rates, before hitting headline unemployment data.

Key Decisions for Investors

• Go long MSFT and NOW on a 6-12 month horizon: both are levered to enterprise AI workflow adoption and seat expansion; risk/reward improves if corporate training spend becomes a budget line item rather than a discretionary HR perk.
• Pair long SNOW / short labor-heavy enterprise software or BPO names on any pullback: the spread favors infrastructure and data-control platforms over firms whose value proposition is mainly labor arbitrage; target 2-3x beta-adjusted upside if AI rollout accelerates.
• Buy calls on PATH or similar automation beneficiaries for a 6-9 month horizon: if employers underinvest in training, they will substitute toward automation; the convexity is strong, but the thesis needs evidence of management teams emphasizing headcount efficiency in earnings calls.
• Avoid or underweight consumer-facing, low-margin service employers with high manual admin content over the next 12 months: these names face a slow-burn margin squeeze from either retention costs or productivity leakage, even before the unemployment data turns.
• Set a catalyst watchlist for earnings season: any management team mentioning AI training, internal copilots, or reduced backfill should be treated as an early signal for margin upside and potential multiple expansion.