Gen AI paradox
What is Gen AI paradox?
The gap McKinsey named in 2025: almost every company now uses generative AI, yet about the same share reports no real earnings impact. The cause is heavy use of broad, low-value tools like chatbots while the function-specific uses that would move the numbers stay stuck in pilots. McKinsey's proposed way out is agentic AI.
Why it matters
Many leaders feel this frustration before they can name it. AI is visibly everywhere in the company, yet nothing lands in the results. The mechanism is subtle. Tools any individual can adopt without asking permission spread on their own, but the time they save is scattered in small amounts across many people, and scattered minutes rarely reassemble into a figure a CFO can see. The uses that would move the numbers run the other way. They cross department lines and need whole processes rebuilt around them, so they stall wherever no one senior owns the change. The practical lesson is that breadth of adoption is a weak proxy for impact. Counting how many people use AI tells you little about whether it pays.
In practice
A company surveys its staff and finds nearly everyone uses an AI assistant weekly, and takes that as success. A year on, no business metric has shifted. The reason is that all that use is individual and optional, saving scattered minutes, while the one workflow that could have cut real cost was never rebuilt, because it needed three departments to agree and no one led the effort.
