McKinsey's State of AI survey found that 88% of organizations now use AI in at least one business function. Only 39% report any EBIT impact at the enterprise level. That gap is the story, not the adoption number.

Most companies can say they "use AI." Far fewer can point to profit that moved because of it. For operators, that should reset the question from "which tool should we buy?" to "which workflow actually changed?"

Source: McKinsey, The State of AI.

Adoption is easy. Impact is not.

Buying seats for a chatbot, turning on a built-in assistant in your CRM, or letting the team draft emails with generative AI counts as adoption. It shows up in surveys. It does not automatically show up in EBIT.

Impact shows up when a process finishes differently: fewer hours on the same work, faster cycle time, fewer dropped handoffs, or revenue that would have leaked without a reliable sequence. That requires changing how work moves between systems, not just adding a model on top of the same manual path.

Why so many deployments stall at "useful but not valuable"

Teams get stuck for predictable reasons:

  • AI sits beside the workflow instead of inside it. People still copy fields, chase reminders, and rebuild reports by hand.
  • Old steps stay in place. You gain minutes on a draft and lose them reconciling the same spreadsheet.
  • Success is measured as usage, not as dollars or hours returned.
  • Pilots stay in one department. Nothing connects intake, follow-up, billing, and scheduling into one owned sequence.

McKinsey's split between widespread use and limited enterprise EBIT impact matches what we see with smaller operators too. The tool is present. The process is still human glue.

What the operators capturing returns do differently

They start with a painful, repeatable job and redesign it end to end:

  • Trigger: form submitted, invoice overdue, appointment booked, quote sent
  • Actions: update the record, send the right message, create the folder or task, notify the right person
  • Stop rules: halt when payment lands, when the client replies, or when a human must decide

That is not an "AI strategy deck." It is a workflow with AI and automation where judgment is not required, and people where it is. The model is optional. The redesigned sequence is not.

A practical test for your next AI spend

Before you add another tool, ask three questions:

  • Which weekly process still needs a person to carry data between systems?
  • If that process ran without them, what hour or dollar number would move?
  • Can you measure that number in 30 days, not in a year-long transformation?

If you cannot answer those, more adoption will not close the 88% versus 39% gap for you. It will widen the list of tools your team half-uses.

AI usage is common. Profit from AI is still rare. The difference is whether the work changed.