McKinsey & Company, "How to close the agentic adoption gap" (August 7, 2026)

On the seventh of this month McKinsey, a firm that sells the very transformation work it is describing, published a finding with a number in it that should have stopped more people than it did. In the AI efforts that actually pay off, for every dollar spent on the technology itself, three go to redesigning how the work happens and five go to building the skills and habits that make the redesign hold. Most companies invert the whole arrangement. The money and the attention pile onto the part you can purchase and count, and the human work that creates the value gets filed under implementation. Then, McKinsey notes, they track log-ins and prompt volume rather than whether the work got any better.

Read that against this week's five pieces and it stops looking like an AI story. Every one of them describes an organization that already holds the knowledge it needs and has built a machine that pays its people not to use it. One article follows a company growing more precise about its own dysfunction and no closer to fixing it. Another lands on a manager whose scorecard pays him to keep his strongest analyst exactly where she is. A third finds a strategy everyone can recite that stopped being true months ago. The last watches a new executive dropped into the hardest seat in the building while a graduate two floors down is handed a mentor and a plan. The information was never the scarce thing. What went missing was any willingness to change the operating system that keeps producing the behavior.

Buying the tool is the easy tax to pay, because it shows up as progress on a slide. The work that would actually move the number rarely does.