AI Adoption Hit 89%. Real Payoff Stuck at 6%, McKinsey Finds

Eighty-nine percent of organizations now use AI in at least one business function. Yet only 6 percent qualify as high performers — companies that attribute at least 5 percent of EBIT to AI and report value from its use — and that figure has not moved in a year, according to McKinsey's latest State of AI research.

The gap is not closing. Thirty-seven percent of organizations report some positive effect on earnings, McKinsey found, but broad adoption is still not translating into legitimate payoff. A senior McKinsey partner, writing in a Fortune commentary, argues that leaders keep mistaking some of AI's clearest successes for a playbook they can apply anywhere.

The evidence executives cite most is real. A study of roughly 5,200 customer-support agents found AI assistance increased issues resolved per hour by 15 percent. Randomized trials involving about 4,900 software developers found those given an AI coding assistant completed roughly 26 percent more tasks. Both gains are meaningful. Neither, the partner writes, establishes that adding AI to any business process will improve a company's earnings.

Contact centers and software teams had structural advantages before today's AI models arrived. Contact centers had spent decades organizing high volumes of work into queues, tracking outcomes, and building a body of past interactions. Software teams had testing, continuous integration, and code review practices that made it possible to inspect and correct new work. AI entered those settings and made an existing system faster — a different challenge from redesigning work that has never been organized around a clear outcome or a way to check whether it was achieved.

The partner points to a bank using AI to read documents for small-business loans. Faster document review might save two days. But if an application still waits on separate handoffs among sales, credit, compliance, and operations, the customer sees little improvement. Genuine redesign would start with the decision itself: what evidence is required to approve a sound loan, who holds authority, which exceptions need specialist review. The bank would then measure time to decision, error rates, and the loans it can responsibly serve — rather than counting only hours saved reading files.

Those choices — about authority, risk, and what employees do with regained time — cannot be made by software. The partner also notes that employees notice the uncertainty and ask, "Am I training my replacement?" Managers cannot offer much reassurance if leadership has described only the tasks AI might perform, not the work people will do next.

Management consulting bears some responsibility, the partner concedes. Consultants often earn their keep by removing waste from an existing process — optimization — and are only now rethinking the purpose of accumulated organizational processes and value tradeoffs.

The numbers support the case for redesign. McKinsey research published in July found that among leaders reporting on organizations in the earliest stage of AI adoption, those whose workflows had been redesigned were 5.3 times as likely to report enterprise-level value as those whose workflows had not: 32 percent versus 6 percent.

That finding should change the first question a CEO asks, the partner writes. Before choosing a tool, decide what outcome the company needs and which parts of the work no longer serve it. The contact-center and coding gains are genuine, but they are a starting point, not a template. The larger opportunity belongs to leaders willing to change the work around the technology.