Apollo Global Management’s chief economist Torsten Slok published a warning this week that should be required reading for every business leader with AI projects on the roadmap: the productivity revolution is real, but it is happening in a very small part of the economy, and the rest of the market may be pricing in gains that have not arrived.

Published July 6 in Fortune, Slok’s analysis points to a striking divergence. Profit margins for the Magnificent Seven (the major US tech companies driving AI infrastructure spending) have risen from roughly 15% to 25% between the first quarters of 2023 and 2026. Margins for the remaining 493 companies in the S&P 500 have stayed flat at around 10%.

That gap is the story. AI is working extremely well for companies that build software and can iterate on digital workflows. It is working much more slowly for companies that still rely on physical processes, face-to-face service delivery, or legacy infrastructure.

The Data Behind the Warning

Slok is not alone in flagging this pattern. A study from MIT found that only 5% of companies saw a meaningful return on investment from generative AI pilot projects. The other 95% either saw no measurable impact or are still waiting.

BCG’s 2026 Global AI at Work report adds detail to that picture. While 42% of respondents reported saving roughly eight hours a week through AI use, most said they received little to no guidance on what to do with that freed time. Half said they were not using it to complete more strategic work. Time saved that gets absorbed back into low-value activity does not show up as productivity growth.

Slok’s concern is that equity markets have been pricing AI as if the productivity gains were already flowing into corporate earnings across the board. If those gains arrive slowly, or not at all for most companies, there could be what he called a “painful repricing” of AI-related stocks.

Why Most Businesses Have Not Cracked It Yet

There are a few structural reasons why AI productivity is concentrated at the top.

Software companies can change a product by updating code. A manufacturer, a healthcare provider, or a professional services firm has to change how people work, which tools they use, how decisions get made, and often how clients are billed. That is a much longer and harder implementation.

Deploying AI also requires decent data foundations. If your data is siloed, inconsistent, or locked in legacy systems, even the best AI tools cannot do much with it. And most Fortune 500 companies still have substantial data infrastructure debt built up over decades.

There is also the question of change management. Research consistently shows that employees often resist AI tools not because they do not see the potential but because they do not trust the outputs, do not know how to verify them, or worry that demonstrating AI-driven productivity will simply result in headcount cuts without any benefit flowing to them.

What This Means for Business

If you are a business leader looking at this and wondering whether AI is worth the investment, the honest answer is: it depends entirely on how you implement it.

The 5% of companies that saw meaningful returns from their AI pilots did not win by luck. They typically did a few things differently. They started with a specific, well-defined problem rather than a broad AI initiative. They measured outcomes from day one. They built internal capability rather than outsourcing the whole project to a vendor. And they treated the AI rollout as a change management challenge as much as a technology challenge.

The companies in the 95% bucket often made the opposite choices. They ran pilots without clear success metrics. They bought tools without changing the workflows around them. They did not invest in training people to use the output critically.

The repricing risk Slok is flagging is real, but it is also solvable. The productivity gap between tech companies and the rest of the economy is not inevitable. It is a consequence of implementation quality, not technology limitation.

The Roadmap for Closing the Gap

For most businesses, closing the gap means getting a few fundamentals right before scaling AI spend.

Data readiness is the foundation. AI agents need accurate, accessible data to produce reliable outputs. Companies that skip this step find themselves building on sand.

Workflow redesign matters more than tool selection. The question is not which AI platform to use but which business process to redesign first, and how to integrate AI into that process so people can trust and act on the output.

Measurement discipline keeps the investment honest. Define what success looks like before you start. Know which metrics will tell you whether this is working. Check those metrics on a regular cadence and be willing to change course.

Finally, internal capability beats vendor dependency. Companies that have people who understand both the business and the technology are the ones turning AI investment into competitive advantage. That might mean upskilling your existing team, hiring for new skills, or bringing in advisors who can help build that capability internally.

Enterprise DNA works with businesses on exactly these challenges, whether that is building data literacy across the team through EDNA Learn, deploying AI agents that actually move specific metrics through Omni Ops, or working through the strategy with Omni Advisory. The productivity gap is real. The question is which side of it you end up on.

---

Want a second opinion on your AI roadmap before you scale spend? Book a call with Omni Advisory to pressure-test your approach.