Artificial intelligence so far has had almost no impact on the financial results of most American companies in the S&P 500 index, with the exception of the technology sector. Torsten Slok, chief economist at Apollo, stated that outside of technology, the return on AI is effectively zero. According to his data, margins and revenue per employee are growing at the "Magnificent Seven," while the remaining 493 companies show almost unchanged figures. According to Slok's estimate, a broad effect for these companies will likely appear in about two years.
Most of the benefit has gone to infrastructure providers. Microsoft's Azure revenue grew by 48 percent, and Google Cloud's by 103 percent, reflecting demand for cloud computing for AI. The companies buying these technologies, by contrast, are barely seeing any margin gains. Adam Doomy notes that individual savings of employees' time do not add up to a tangible result for the enterprise as a whole: the saved minutes are absorbed by the same old procedures, and the company's overall performance does not change.
Surveys show widespread but superficial use of the technology. About 40 percent of workers use generative AI, and roughly 25 percent of firms have implemented it in their operations. The gap between employees' personal experience and corporate results is the heart of the problem. Key workflows and approval procedures are still built around the human pace of work, so point improvements do not turn into systemic change.
Doomy argues that the problem cannot be handed off to IT departments. Responsibility for the transformation, he says, should be taken on by the CEO, not by technical units, which as a rule are responsible only for deploying tools. He advises starting with essential workflows directly linked to revenue, margin, risk, and cycle times, rather than with scattered experiments. This, in his view, will require revising employee roles, decision-approval authority, and mechanisms for monitoring system performance in real time.
The author also points to the choice of models. Companies use a mix of open and closed models, but many rely excessively on cutting-edge solutions where they are not needed, which increases costs. The BCG index records a rise in AI spending to 3.3 percent of revenue from 1.7 percent previously, nearly double. At the same time, enterprises do not know how to measure the cost of an outcome per unit of product or operation, and dashboards track only the level of technology usage, not its financial effect.
The column also recalls the social side of the issue. On October 22, Barclays cut about 5,000 jobs and cited AI, but it is not specified what share of these cuts is directly related to the technology. Such decisions show that expectations about AI are already influencing hiring policy at banks and other companies, even though measurable returns for most of them are not yet visible. The author expects that noticeable and widespread benefits from AI are unlikely to appear before the end of 2027, and a significant effect for companies in the S&P 493 may emerge only in a couple of years.