What Apollo found
On October 24, 2026, Torsten Slok, the chief economist at Apollo, published a short note titled No Signs of AI Payoff Outside the Tech Sector. He split the S&P 500 into two groups: the seven largest tech companies, often called the Magnificent 7, and the other 493 companies in the index.
For the tech giants, revenue per worker and profit margins have been climbing. For the other 493 companies, both have stayed flat. Apollo also found no sign that AI is lifting productivity growth across the economy. Slok concluded that the AI story will be decided by those 493 companies. They include the banks, manufacturers, retailers, health systems, and service firms that buy AI rather than sell it.
Revenue per worker, inflation adjusted
Magnificent 7 tech companies
Other 493 S&P 500 companies
Apollo also pointed out that AI spending is concentrated in a small number of firms, and that for mainstream companies AI spending so far shows up as a growing cost rather than a source of growth. The results have not reached revenue, margins, or output per employee.
The same gap shows up inside individual companies
Apollo looked at the economy from the top down. MIT looked at individual projects a year earlier and found a similar picture. In its 2025 report on generative AI in business, MIT found that only about 5 percent of AI pilot programs produced rapid revenue gains. Most stalled with little or no measurable effect on profit and loss, according to Fortune's coverage of the report.
MIT traced the shortfall to what it called a learning gap and to flawed integration into how companies actually work, rather than to the quality of the AI models. MIT also found that companies working with specialized vendors and outside partners succeeded about 67 percent of the time, while internal builds succeeded about one third as often.
AI project success rate by approach
What the research does and does not say
Neither study says AI cannot pay off. Apollo's note tracks broad averages across hundreds of companies, so it shows the overall result without explaining what any one company did. MIT's numbers come from interviews, surveys, and public deployments, and they describe pilots, not mature programs. Both point to what we see with our clients. Buying the tools is the easy part, and the payoff comes from changing how the work gets done.
“Most companies we meet already have AI tools in place at their organizations. The payoff depends on what happens after the tools are purchased, and the common mistake is in the change management and rollout. The companies with the highest gains, operationally and financially, are the ones with a plan before they even begin.
Leadership needs to discuss and decide how AI fits in the business, how security is scoped and which policies are required, how teams will be trained, and, ultimately, what the desired outcome is. We work through each of these with our clients so they get the best possible implementation of AI in their organization.”
Five decisions to make before the rollout
With our clients, the rollouts that produce results start with leadership working through five decisions. None of them require a technical background.
Where AI fits in the business
Start with the work where AI should help first, usually repetitive, document-heavy tasks with a clear owner. When AI is pushed into every department at once, usage stays light and very little changes.
How security is scoped and which policies are required
Decide what company and client data can go into which tools, which vendors and models are approved, and what the acceptable-use policy says. Setting these rules before people form habits is far easier than correcting them later.
How each team will be trained
Executives, managers, and front-line staff use AI for different work. Training needs to be built around each group's real tasks so people keep using it after the session ends.
Who owns the rollout
Someone needs to own adoption, answer questions, and check how the tools are being used. Without an owner, usage tends to fade once the launch is over.
What the desired outcome is
Agree on what success looks like and how it will be measured, such as hours saved on a process, faster turnaround for clients, or fewer errors in a report. Without a target, there is no way to tell whether the spending is paying off.
How Blue Brick Partners helps
We work through these decisions with leadership first, then carry them into the rest of the company. That includes private, hands-on training for executives, personalized sessions for employees at every level, the governance and security policies that keep company data safe, and the workflow changes that turn AI use into measurable results. We are based in Dallas, Texas, with a satellite office in Nashville, Tennessee, and work with companies nationwide, on-site or remotely.
See our AI consulting and training services
Common questions
Why are most companies not seeing a return on AI?
Research from Apollo and MIT shows that AI has not yet changed financial results for most companies outside big tech. MIT found that most generative AI pilots stall because the tools are not fitted to company workflows and the people using them are not trained to get value from them. As James Williamson, Managing Principal at Blue Brick Partners, puts it: "The payoff depends on what happens after the tools are purchased. The companies with the highest gains, operationally and financially, are the ones with a plan before they even begin."
Does the Apollo study mean AI does not work for business?
No. Apollo looked at aggregate revenue per worker, profit margins, and productivity across the S&P 500. Those numbers show no payoff yet outside the largest tech companies, but they do not measure individual rollouts. The same data shows the largest tech companies pulling ahead, so the gains are real for some firms.
What should leadership decide before rolling out AI?
Leadership should decide where AI fits in the business, how security is scoped and which usage policies are required, how each team will be trained on its own work, who owns the rollout, and what measurable outcome the company expects. These decisions are easier to make before tools are purchased than after usage habits have formed.