Apollo Global Management's chief economist Torsten Slok published a five-chart analysis on October 24 that landed like a grenade in the AI investment debate: despite massive spending, artificial intelligence has delivered "zero financial payoff to corporate America outside of tech." Revenue per worker is rising for the Magnificent Seven but flat for the other 493 companies in the S&P 500. Profit margins tell the same story. And US productivity data shows no AI-driven boost whatsoever.
The analysis matters because it directly challenges the case that every business owner is currently weighing. If the largest, best-resourced companies in the world cannot show AI moving the needle on their financials, why should a 30-person firm in Brisbane invest? But Slok's provocative headline — while grounded in real data — tells only half the story.
What Slok's charts actually show
Slok's note, covered by Bloomberg the following day, uses five charts built on Bloomberg, Macrobond, Bureau of Labor Statistics, and Ramp data. The first two show inflation-adjusted revenue per employee at roughly $780,000 for the Magnificent Seven versus $525,000 for the S&P 493 — with the gap widening since 2022. Profit margins paint an identical picture: near 25% for Big Tech, approximately 10% for everyone else.
The third chart shows total factor productivity running at its long-run average of 2.1% since 1947 — no AI-driven acceleration. The fourth and fifth, using Ramp's spending data, reveal that the top 10% of firms account for 99.5% of spending on model inference, and 80.4% of the top 1% of AI spenders are venture-backed.
"AI may be a great calculator, but in business, intelligence is mostly judgment," Slok wrote. "Judgment depends on context, and the answer varies with which human or which model you ask."
It is a sharp argument from a serious economist. But aggregate metrics can obscure what is happening at the firm level.
The counter-evidence: a J-curve, not a flatline
Multiple data points suggest Slok's "zero payoff" framing, while accurate at the aggregate level, misses important nuance.
A new academic study of AI adoption across S&P 500 firms found what researchers call a J-curve effect. Non-technology firms in the early stages of AI deployment actually showed 2 to 3 percentage points lower profit margins than firms with no AI adoption. But firms that achieved deep integration — the highest adoption tier — showed 12.6 percentage points higher net profit margins. The lesson: superficial AI adoption can make things worse before deep deployment makes them better.
McKinsey's 2026 State of AI survey reinforces this pattern. While 80% of individual respondents say AI has improved their personal productivity, only 37% of organisations report any positive EBIT impact — unchanged from 2025. Just 6% qualify as "high performers" with at least 5% of earnings attributable to AI. The distinguishing factor? High performers "fundamentally redesign workflows that are enabled by AI rather than insert AI into existing ones."
Bain's 2026 automation survey puts a number on the gap: enterprises are realising just 38% of AI's expected value. Nearly 40% of companies that measured AI cost savings landed below 10% of their targets. Only 7% run fully autonomous agents in production, yet many investment cases assume full autonomy economics.
Even where gains are real, they are not always visible in the numbers Slok is measuring. J.P. Morgan Asset Management noted that American Express cut coding cycle times by more than 30% but is using the extra capacity to tackle a project backlog rather than reduce costs. Netflix is producing content faster and more cheaply but reinvesting the savings into more content. These are genuine productivity gains that never show up in revenue-per-worker charts.
The Australian context makes this personal
This debate has particular weight in Australia. The Tech Council of Australia's October report found that only 26% of Australian SMEs are using AI at all, and just 5% have adopted it broadly across their operations. The report estimates that an average SME moving from no AI to broad adoption could unlock $136,000 in annual productivity gains — a 36% uplift.
That aligns with the J-curve finding. Australian SMEs are overwhelmingly in the "not yet adopted" or "superficial adoption" categories — precisely the zone where Slok's charts show nothing happening. The businesses that have gone deep, like Wesfarmers deploying AI agents across Bunnings, Kmart, and Officeworks to save 500,000 hours, are the exception, not the rule.
The risk for Australian businesses is reading Slok's headline as permission to wait. As we have previously reported, Australia already ranks last among comparable markets on deep AI integration, with just 10% of businesses having embedded AI into their operations. A pattern across OECD data and now Apollo's analysis is becoming clear: surface-level AI adoption produces surface-level results.
What to watch
Slok is right that the AI story "will be decided in the S&P 493, not the Mag 7." But the evidence increasingly suggests the deciding factor is not whether AI works — it is whether organisations are willing to redesign their workflows around it, rather than bolting it onto processes designed for humans.
For business owners weighing AI investment, the practical takeaways are sharper than "AI doesn't work" or "AI is essential":
- Measure at the process level, not the program level. Bain's data shows realization rates vary wildly by function. A customer service deployment might deliver 80% of expected value while a marketing one delivers 15%. Aggregate ROI is meaningless without process-level attribution.
- Budget for the J-curve. The academic evidence says early-stage adoption often makes margins worse before deep integration makes them better. If you are six months in and seeing costs rise without returns, that might be the expected path, not a failure signal.
- Watch for reinvested gains. If your team is using freed-up time to tackle the project backlog or serve more customers, AI is working — it is just not showing up where you are measuring.
The companies that burned through AI budgets without a strategy and those that pulled AI tools after superficial deployments are cautionary tales on both sides. Apollo's data does not prove AI is failing. It proves that most companies have not yet done the hard work of making it succeed.
Sources
- No Signs of AI Payoff Outside the Tech Sector — Apollo Global Management, Daily Spark
- AI Provides 'Zero Payoff' to S&P 493 Outside Tech, Apollo Economist Says — Bloomberg
- AI payoff outside tech: S&P 493 margins flat, Apollo says — BullQuill
- AI Adoption in S&P 500 Firms — arXiv
- The State of AI: Global Survey 2026 — McKinsey & Company
- Realization Rate: The Metric Reshaping Enterprise AI — Bain & Company
- Are Companies Benefiting from AI Adoption? — J.P. Morgan Asset Management
- Accelerating AI Impact for Small and Medium Enterprises — Tech Council of Australia
- AI Could Add $93 Billion to Australian Small Businesses — Google Australia Blog
