Reaction · October 5, 2026 · Far West Consulting
The most popular AI use cases are rarely the ones that pay back
Gartner's September survey of 1,303 leaders found only 22% have scaled AI, and the use cases leaders pursue most are rarely the ones reporting the best returns. We read it as an argument for choosing from the workflow.
For CFOs, COOs and functional leaders deciding where the next AI dollar goes.
On September 1, Gartner reported on a survey of 1,303 leaders at organizations with at least US$50 million in annual revenue, fielded January to April 2026.20 Only 22% had scaled AI across multiple business units or adopted an AI-first approach. About 11% could not say what their function spent on AI in 2025. And 85% of functional leaders plan to spend more in 2026, after putting an average of 12% of their functional budgets into AI in 2025.
Gartner also sorted respondents into high and low performers. The high performers constantly track the return on their AI initiatives, treat AI as a portfolio, and regularly reallocate or discontinue what underperforms. They reported positive returns on 81% of their initiatives. The low performers couldn’t say what the return was on 29% of theirs. Gartner doesn’t say these are the same companies as the 22% that scaled. They’re two cuts of one survey.
The more useful finding is about what people chose to build. Gartner says the most widely pursued AI use cases are rarely the ones with the largest share of leaders reporting positive returns. Its example is IT: threat detection and service-desk automation were the most-pursued use cases (54% each), while IT asset and cost optimization topped the positive-returns list (40%). It’s one function’s example, and it counts leaders reporting a positive return, not dollars.
What we make of it: a use case gets popular because it demos well. It pays back because it fits how a particular team works. Those aren’t the same list, and Gartner’s data shows the gap. It’s the order we work in: map the team’s actual work first, then choose the tool, rather than the reverse. The AI Workflow Discovery engagement exists to do that first step before anyone commits to a platform.
The 11% matters as much as the 22%. A function that can’t say what it spent can’t say what it got back, and can’t move the money when a project stalls.
A use case nobody can measure is a bet. Choosing from the work makes it a smaller one.