New AI Jobs Index Ranks 784 Occupations By Loss Risk

Median scenario: 9.3 million U.S. jobs projected at risk from AI (range 2.7M–19.5M) with ~$757 billion in annual income exposed. Occupations with highest projected risk include writers/authors (57%), computer programmers and web/digital designers (55%), editors (54%), and web developers (46%); sectors most vulnerable: Information (18%), Finance & Insurance (16%), and Professional, Scientific & Technical Services (16%). Authors highlight an "augmentation-displacement" link where AI-driven productivity can reduce hiring, and note job-creation effects, regulatory constraints, and union impacts are not yet included in the index.

Analysis

The key investment dynamic is not just automation of tasks but a structural reallocation of labor budgets: firms that deploy LLMs to augment entry-level productivity can satisfy throughput targets with fewer hires, converting headcount growth into margin expansion. Expect the largest P&L effects to show up in operating expense lines (SG&A and hiring budgets) within 6–24 months as pilots scale into procurement and hiring freezes replace layoffs initially.

Second-order winners are vendors that sell observability, governance, and human-in-the-loop tooling — the rise of AI ops and content-moderation layers will create a new, sticky software TAM that commands enterprise spend above pure compute. Conversely, businesses that monetize volume of low-cost creative labor (staffing firms, local media classified businesses, early-career placement platforms) will face secular headwinds; the revenue compression will compound when client budgets shift from per-head billing to platform subscriptions.

Regulatory and labor counterforces are the most important catalysts to watch: unionization, licensing rules, and procurement restrictions can slow displacement for 12–36 months, creating episodic windows where tech winners' valuations decouple from real adoption. Also note the index omits job creation — roles in AI productization, fine-tuning, safety, and compliance are nascent but likely to absorb some displaced workers over a multi-year horizon, muting net employment loss versus headline projections.

For portfolio construction, the practical path is to overweight infrastructure and enterprise AI tooling while hedging cyclical consumer exposures tied to entry-level wages; size positions to allow for lumpy adoption and policy risk, and prefer option structures to buy convexity to adoption inflection points.

Key Decisions for Investors

• Long NVDA (12–24 months): buy 2026–2027 LEAPS call spreads to capture continued AI compute demand while capping premium. Risk: GPU cycle disappointment / capex pullback; Reward: >2x on sustained data-center orders.
• Long SSTK (6–12 months) vs Short MAN (ManpowerGroup) (6–12 months) pair: overweight Shutterstock (equity or 9–12 month calls) to play higher-margin AI-assisted content products; short staffing firms exposed to entry-level placement to capture revenue erosion. Risk: faster-than-expected licensing competition for SSTK; Reward: asymmetric if content platforms monetize AI tooling.
• Long Coursera (COUR) or other reskilling plays (9–18 months): buy equity or call spreads to capture corporate training budgets shifting to AI reskilling. Risk: weak monetization / competition from free models; Reward: durable ARPU expansion as firms react to displacement costs.
• Tail hedge: buy 3–6 month puts on a concentrated AI hardware/software long (e.g., NVDA or MSFT) sized to limit portfolio drawdown from regulatory shocks or a sudden enterprise pause in AI adoption.