Apollo Economist Says AI Has Yet to Pay Off for Most Non-Tech Companies

Introduction

Artificial intelligence has moved quickly from an experimental technology to a standard item on corporate budgets and strategy documents. Yet the financial results have not necessarily followed the spending. In a recent note, Torsten Slok, chief economist at Apollo Global Management, argued that AI has produced no clear financial payoff for large U.S. companies outside the technology sector.

Key points

• A narrower market benchmark: Slok focuses on what he calls the “S&P 493,” meaning the S&P 500 minus Alphabet, Amazon, Apple, Meta, Microsoft, Nvidia and Tesla. Removing these companies is intended to limit the influence of firms that build or sell much of the AI ecosystem.
• No visible improvement in major indicators: Profit margins and revenue per worker for the broader non-tech group appear flat, with no evidence that AI is lifting productivity at the company level.
• Cheap intelligence is not automatically new revenue: Slok is skeptical that businesses can easily turn falling AI costs into faster revenue growth or stronger margins.
• Business value involves more than computation: AI can calculate, summarize and generate outputs, but commercial decisions also depend on judgment, accountability, trust, culture, unwritten rules, ethical trade-offs and fairness.

Why adoption has not translated into returns

The first financial effect of AI adoption often appears as software purchases, pilot programs or isolated workflow improvements. Those changes may save time for a team without changing a company’s total revenue or cost structure. Unless freed capacity becomes additional sales, higher output or durable cost savings, local efficiency will not necessarily show up in the income statement.

Many business processes also depend on context that is difficult to formalize. Customer relationships, brand trust, risk ownership and cross-functional coordination cannot be resolved simply by making information processing faster or inference cheaper. If AI-generated work requires extensive review, approval and accountability checks, some of the expected savings can be offset by new oversight costs.

Implications

The analysis should not be read as proof that AI can never create value. It is a warning to separate three different claims: that companies are adopting AI, that a particular team is becoming more efficient, and that the company as a whole is earning more or spending less. Technology companies are both major suppliers of AI infrastructure and among its most direct beneficiaries, so their performance should not automatically be treated as a picture of the wider economy.

For non-tech businesses, the next test is likely to be less about how many models they deploy and more about whether AI consistently improves revenue, costs, delivery quality and decision-making. Only when those effects persist in financial results will the current AI investment cycle become broad productivity growth rather than a story about spending and expectations.

Source: TechCrunch AI