The corporate enthusiasm surrounding generative language models is giving way to rigorous financial scrutiny. Empirical surveys published by McKinsey & Company and KPMG in early September 2026 indicate a structural shift in enterprise artificial intelligence. Rather than pouring capital into fragmented pilot programs, executives are concentrating on autonomous AI agents, operational inference costs, and deep workflow restructuring.
McKinsey's global study, surveying 1,719 executives across 97 countries, points to a pronounced productivity paradox. An impressive 80 percent of surveyed employees report that AI tools have enhanced their individual productivity, with half stating that it improves their decision-making. Yet at the organizational level, this lift barely surfaces: only 37 percent of companies can demonstrate a measurable positive impact on corporate operating earnings (EBIT), roughly unchanged from the previous year. Just six percent qualify as high performers capturing more than a five percent EBIT contribution.
At the same time, the research challenges widespread fears of immediate, massive workforce reductions. In 2025, 32 percent of surveyed organizations expected AI-driven headcount cuts within twelve months. Looking back from 2026, only 14 percent actually carried out layoffs attributable to AI, while roughly 67 percent recorded no staffing changes whatsoever. For the coming year, 39 percent anticipate a modest reduction in headcount, 43 percent project flat staffing, and ten percent plan to hire additional personnel.
Specialized agent systems are profoundly altering internal software engineering and IT procurement. McKinsey found that 40 percent of large enterprises with over one billion dollars in revenue are now scaling AI agents, compared to 27 percent a year earlier. For small and mid-sized enterprises, adoption remains stalled at 22 percent. Furthermore, 31 percent of large corporations are scaling agentic software development tools, and 32 percent of all firms have actively decided against purchasing external software licenses, choosing instead to build custom modules in-house using coding agents.
A complementary report by KPMG surveying more than 1,000 chief financial officers and risk executives illustrates a parallel acceleration in the financial sector. The proportion of firms actively deploying and scaling AI within their finance function jumped from 30 percent in 2024 to 75 percent in 2026. Fully 77 percent of financial institutions have moved past conceptual planning for autonomous agents into active testing or production environments, with 71 percent reporting faster decision-making and 64 percent seeing greater forecast accuracy.
Sustained returns, however, appear tightly bound to operational oversight and cost management. KPMG notes that companies with formal governance and assurance mechanisms achieve a 33 percent error reduction, compared to just six percent among laggards lacking structured controls. Meanwhile, cost pressures are mounting: McKinsey reports that escalating day-to-day token and inference expenses have prompted 20 percent of enterprises to cap the expansion of additional AI use cases.