AI investments increase but results are unclear... McKinsey: "Management issues are more significant than technology"

Although corporate AI investments are increasing rapidly, it appears that not many companies answered that this has actually led to improved performance. The diagnosis offered at the Google Cloud Next 2026 event was clear. For today’s enterprise AI, the core challenge is not the act of adopting the technology itself, but rather “who takes responsibility and how the organization is mobilized.”

McKinsey’s Sutoshi Padhi(Asutosh Padhi), a senior partner, said that recent surveys assess that about 90% of AI projects fail to create clear business value. He explained that CEOs( and CFOs) are commonly sharing their concern that “IT spending continues to rise, but returns are not at all clear.”

Padhi viewed this issue as starting not from simple technological limitations, but from a lack of ownership among executives. The moment AI is no longer a central agenda in the enterprise strategy and is delegated to the CIO( or to data analytics leaders, it becomes hard to create performance. He pointed out that in organizations where, when the CEO or CFO asks about the progress of AI, people divert attention to other executives, the organization has effectively already lost direction.

According to McKinsey’s research, only about 39% of companies said they experienced a real increase in profit from AI investments. This means that most companies have not yet reached “enterprise-wide performance.” There is also a complex, intertwined data environment behind this. The explanation is that enterprise resource planning) (ERP) systems introduced at different times, data that was not integrated after mergers and acquisitions, and information structures fragmented by department are blocking the use of AI.

Start with core issues, not easy tasks

Padhi identified a mistake many companies make here: taking an “approach that starts with easy tasks.” Even if dozens of small pilots that seem to deliver quick results are run, it is difficult to turn them into meaningful change unless they spread across the entire organization. Instead, he emphasized that solving the most difficult and important business problems first concentrates the organization’s attention and resources, and that change management and capability building follow along.

He said that resolving issues that directly affect corporate value is what makes everyone pay attention first. Meanwhile, he diagnosed that even if simple use cases succeed, they are likely to be treated inside the organization as “side experiments,” making it hard to gain momentum for expansion. This is the interpretation of why, despite many AI investments, performance looks blurred.

McKinsey’s solution and leadership change

McKinsey proposed an “AI management operating system” as its solution. This is an operating model that embeds AI into day-to-day work flows within a structure that stays connected from the CEO to frontline practitioners, enabling faster decision-making. In particular, Padhi claimed that companies that implement their core business structure in a highly precise digital twin form can shorten new product launch cycles by 70% or more.

He also expected that leadership criteria in the future will change significantly. First, he stressed that executives must understand the technology themselves, and it should not be handed off only to external parties or operating departments. In addition, he said that execution speed and human judgment are becoming more important. Although information can be obtained more easily, how that information is used and what decisions are made ultimately depend on uniquely human capabilities such as empathy, kindness, and judgment.

This remark is drawing even more attention because it comes at a time when Google Cloud is moving to fully expand its “agent-based infrastructure.” In the market, people say that while competition in AI investment will continue to heat up, the real contest is increasingly likely to be decided by organizational operations and leadership rather than by model performance. Ultimately, the real question in this technology cycle is becoming less about “whether AI has been adopted” and more about “whether AI has been made into a management system.”

TP AI 유의사항 TokenPost.ai 기반 언어 모델을 사용하여 기사를 요약했습니다. 본문의 주요 내용이 제외되거나 사실과 다를 수 있습니다.