Spending on AI in operations has risen sharply over the past three years. Utilities, manufacturers and hospital systems are applying it to maintenance, planning, quality and scheduling, and most operations leaders now count AI among their top three priorities. Yet few can point to its effect on the numbers that matter: cost, output, uptime and working capital.
This benchmark measures that gap. We surveyed 286 heads of operations, manufacturing and asset management, and visited 40 sites to see how AI is used in practice. The report identifies a small group of companies, about one in six, that have moved AI into daily operations at scale, and describes what they do differently from the rest.
The difference is rarely the technology. Leaders and laggards use similar tools and often similar models. They differ in how they choose use cases, how much they invest in changing work around the model, and how they measure results.
Key findings
Most use cases stall. Of the operations AI use cases respondents had launched since 2023, 72 percent are still in pilot or have been stopped. Only 11 percent run across all relevant sites.
Few see it in the accounts. 18 percent of companies say AI savings are visible in their reported operating costs. Most measure success by model accuracy or user adoption.
Leaders invest in the work, not the model. Companies with scaled AI spend about three dollars on process change, training and system integration for every dollar spent on building models. Laggards spend roughly the reverse.
Maintenance leads, planning follows. Predictive maintenance is the most commonly scaled use case, followed by demand and production planning. Quality inspection has the highest share of stalled pilots.
Trust is earned with explanations. At sites where AI output is accompanied by its reasons, frontline adoption rates are more than twice as high as where it is not.
Data is a smaller barrier than assumed. Respondents ranked data quality as their top obstacle, but among companies that scaled AI, the most cited obstacle had been changing planning routines.
Methodology
The survey was conducted between May 12 and June 27, 2026, among 286 senior operations executives at companies with annual revenue above $1 billion in energy and utilities, industrials, healthcare and consumer goods, across North America, Europe and Asia-Pacific. Respondents reported on a total of 1,940 AI use cases.
Companies were classified as leaders if at least three AI use cases ran at scale across their operations and produced savings confirmed by their finance function. Forty site visits, including interviews with planners, supervisors and frontline staff, were carried out between June and August 2026.
