AI adoption looks strong at first glance, and familiarity with tools such as ChatGPT or Copilot is no longer a barrier. Yet three 2025 studies find that everyday use seldom turns into measurable business value. Read together, they describe a funnel that narrows sharply from top to bottom.
At the top of the funnel, Wharton’s 2025 adoption study found that 82% of business leaders use generative AI tools weekly, and nearly half use them daily. Three in four report some form of positive return. AI has become a normal part of the working day across most business functions.
The middle is where the funnel thins. MIT’s GenAI Divide study found that only 5% of companies’ AI pilots make it into everyday operations, and most show no measurable effect on profit. Off-the-shelf tools even outperform custom-built ones, so the weak point is how the tools fit into the work. Many projects never learn from the context they run in, connect to core systems or work across teams. That is why pilots stall while daily use looks healthy.
At the bottom, impact across the whole company is rare. McKinsey’s State of AI 2025 study finds that 8 in 10 organisations use generative AI in at least one function. Only 39% report any effect on enterprise-wide EBIT. The organisations that do scale have redesigned their workflows and operating models, so that AI works as a capability across the whole business.
Value takes more than use. AI has to be built into how data moves through a company and who makes which decisions. Doing so takes spending on system integration, governance and training. High-performing firms treat AI as digital labour, a core part of how work is done. That means connecting it to the systems that hold a company’s official records and redesigning roles to remove manual steps. It also means tracking where value is created and giving someone clear ownership of AI in operations.
Most organisations are stuck in the middle of the funnel. Their people use AI every day, but the organisation does not change around them, so productivity gains stay with individual teams and never reach the enterprise.
