Vanguard Chief Economist: AI and jobs, still in an ATM phase

The article argues that fears of widespread AI-driven job losses look overblown, drawing a parallel to ATMs in the 1980s: while routine tasks were automated, teller employment stayed broadly stable from 1980 to 2010 as banks expanded branch economics. It claims the bigger employment disruption came only later with mobile banking, when branch reliance fell from 36% (2007) to 9% (by 2025) and teller employment declined. By analogy, the piece suggests today’s large language models may be closer to the “ATM phase” (augmentation) than the “mobile banking phase” (restructuring), implying limited near-term labor-market disruption in AI-intensive occupations.

Analysis

The market is likely still pricing AI as a near-term headcount shock, but the more investable mechanism is workflow redesign: vendors that sit inside operating systems, ticketing, payments, compliance, and customer service should capture budget before labor savings show up in P&L. That argues for owning the picks-and-shovels of process automation rather than betting on an abrupt collapse in white-collar employment; the lag from pilot to enterprise-wide redesign is usually measured in quarters, not weeks.

For banks, the first-order effect is not fewer employees, but a mix shift toward higher-value roles and better unit economics. That is constructive for large platforms with distribution and data scale, but less so for branch-dependent regionals that need volume growth to offset fixed costs. The second-order winner set likely includes cybersecurity, fraud, and identity vendors because more digital workflows increase attack surface and compliance load.

The contrarian risk is that consensus may be overfitting to the wrong labor signal: if AI does not visibly destroy jobs, wage pressure may stay sticky, which can keep services inflation and SG&A elevated longer than bulls expect. The real falsifier for the "AI augments first" thesis is not an abstract labor report; it is evidence that enterprise software budgets are converting into net headcount reduction in the next 2-4 quarters, or that banks materially cut branch and operations expense ratios without a corresponding rise in digital acquisition costs.

Net: this is not a high-conviction event trade, but it does support a selective long basket of workflow enablers and digital banking infrastructure, while fading the most aggressive anti-labor narratives. The move in labor-sensitive names looks more like a multi-quarter debate than an immediate earnings shock, so timing should be disciplined around guidance updates rather than headline AI sentiment.

Key Decisions for Investors

• Initiate a 6-18 month long basket in NOW / MSFT / INTU on pullbacks of 5-8%: these names monetize workflow redesign and embedded automation, which is the faster path to ROI than outright job destruction; target 15-20% upside with 10-12% downside if enterprise IT spend rolls over.
• Buy JPM vs. short KRE as a 3-6 month pair: the largest banks can absorb digital adoption and convert it into fee density, while branch-heavy regionals are more exposed if mobile/AI compresses local customer traffic; stop if KRE outperforms JPM by >7% on a sustained credit or rate rally.
• Add PANW or CRWD on weakness as an AI-enablement hedge for 6-12 months: deeper digitization expands the attack surface and compliance burden, creating a second-order beneficiary even if labor displacement stays muted; risk/reward is attractive if AI spend broadens beyond model training into production workflows.
• Avoid or reduce short exposure to staffing and outsourcing names (MAN, RHI, KFY) until earnings prove AI is reducing requisition volume; near-term labor data suggests the feared displacement is not yet broad enough to justify a violent multiple compression.
• Set an alert on enterprise software commentary for net headcount reduction and seat-velocity metrics: if management teams start quantifying 3-5% workforce cuts tied directly to software deployments, the thesis flips toward a more aggressive automation short basket.