Use of Gen AI in the Workplace and the Value of Access to Training

The November 2025 SCE finds 39% of employed respondents have used AI tools at work in the last 12 months, with adoption concentrated among college graduates (58.7% vs. 22.9% for non-graduates), high-income workers (66.3% above $200,000 vs. 15.9% below $50,000), and full-time employees (42.7% vs. 24.7%). Workers value AI training, but employer provision is limited: 38% say training is important, while only 15.9% report access to employer-offered AI training. Among those without training, average willingness to pay is 11.4% of salary; among those with access, the required salary premium to give it up is 24.2%.

Analysis

The key market implication is not that AI is “being adopted,” but that adoption is becoming a labor-market sorting mechanism. High-income, high-education, full-time roles are pulling away first, which means the productivity uplift is likely to accrue disproportionately to firms with white-collar task mix and formal training budgets, while lower-wage service and operational employers risk a widening unit-cost gap if they cannot standardize AI workflows. That should create a second-order margin divergence between software/knowledge-work-heavy employers and labor-intensive sectors over the next 12-24 months.

The training data matters more than the usage data because it signals a conversion problem: many workers see the tool, but only a minority are being trained to turn access into measurable output gains. That suggests the near-term monetization layer is not consumer chat usage; it is enterprise enablement, workflow integration, and compliance/admin tooling. The biggest beneficiaries are likely to be firms selling implementation, model governance, security, and role-specific copilots, while generic “AI feature” announcements without training/rollout infrastructure may disappoint as usage stalls after initial curiosity.

The contrarian read is that the market may be overestimating near-term employment disruption and underestimating near-term wage polarization. Workers most exposed to AI also appear most likely to value training, implying AI is still a complement to labor rather than a clean substitute; firms that invest in training can capture productivity without immediate headcount compression. The tail risk is policy backlash or internal resistance if employers use AI to intensify work without sharing gains, which could slow adoption over the next several quarters and cap the upside for AI-exposed software names.

For investors, the setup favors a barbell: long enterprise enablers with measurable workflow penetration, short or underweight broad labor-displacement beneficiaries priced for rapid automation. The next catalyst is not a macro print but earnings season commentary on training spend, seat expansion, and AI-driven productivity metrics; those disclosures will separate real operating leverage from marketing noise.

Key Decisions for Investors

• Long MSFT / SNOW on a 3-6 month horizon: both are better positioned to monetize enterprise AI through workflow integration and governance rather than simple model exposure; upside is strongest if management quantifies training-linked seat expansion and retention gains.
• Long DUOL? No — instead long PLTR vs. short generic IT services basket (e.g., ACN/IBM proxy) over 6-12 months: PLTR is more levered to operational deployment, while large services firms face margin pressure if clients shift spend from labor to software without expanding total budgets.
• Buy 6-9 month call spreads on MSFT or NOW into earnings: the market is likely underpricing incremental ARPU from AI-enabled productivity tooling, but cap upside with spreads because adoption evidence is real yet still early-stage.
• Short a basket of labor-intensive, low-margin employers with limited training capacity (select retail/restaurant or BPO names) versus long software automation enablers over 12 months; thesis is widening unit-cost dispersion, not immediate job loss.
• Avoid chasing pure-play AI hype names that rely on rapid workforce replacement narratives; the data imply enterprise buyers care more about training and integration than substitution, which favors slower, stickier monetization and lowers the odds of a near-term step-function in revenue.