Apollo data shows artificial intelligence yields no productivity gains outside tech

A fresh analysis of company financial performance by investment firm Apollo Global Management shows that artificial intelligence has yet to generate measurable gains in productivity or profitability for ordinary commercial enterprises.

Apollo chief economist Torsten Slok evaluated performance across United States corporate filings, contrasting the largest tech platforms with the broader market. While revenue per worker and profit margins have climbed steadily for the major technology corporations, both indicators have remained flat across the remaining firms in the S&P 500.

Overall productivity statistics show a similar disconnect. Total factor productivity, which measures how efficiently businesses combine labor and capital to create output, has stayed flat despite hundreds of billions in corporate spending on software subscriptions, hardware and computational tokens.

The gap between tools and business judgment

Slok noted that while software vendors market algorithms as operational replacements for staff, business operations depend heavily on context and human judgment. Software can calculate numbers or draft text, but cannot manage accountability, company trust, ethical trade-offs or workplace culture.

The research indicates that mainstream companies are starting to reconsider uncontrolled AI spending. Rather than treating machine learning integrations as an automatic engine for company growth, many finance leaders now view token and subscription costs as an overhead expense that eats into margins without improving output. Most businesses outside technology are declining to pay premium rates for the newest experimental models, preferring to limit spending to basic office automation.

What this means for Australian businesses

The findings offer a reality check for Australian owners facing pressure to adopt every new workplace AI subscription:

• Audit current software add-ons to identify tools that fail to cut hours or improve completed work.
• Focus automation on specific clerical tasks, such as calendar bookings, basic data cleaning and initial customer routing, rather than delegating complex decision-making.
• Avoid paying for expensive top-tier model tiers when smaller, task-specific software performs identical everyday office jobs at a fraction of the cost.
• Treat AI services as an operating expense that requires direct proof of cost savings before renewing annual licences.