Harness finds AI costs outpace governance in firms

Harness has published its 2026 State of AI in FinOps report, which highlights gaps in how companies track and govern rising AI spending.

The study surveyed 700 engineering leaders and practitioners across the United States, the United Kingdom, France, Germany and India. All worked at organisations that actively use AI or large language model services, spanning software engineering, development, DevOps, IT operations and executive leadership.

Its central finding is that AI spending is spreading across infrastructure, software and model costs faster than many businesses can track. More than half of respondents, 52%, said there is no clear owner for AI costs in their organisation, with responsibility split across engineering, FinOps, finance and IT.

That lack of ownership appears to be driving cost shocks. Some 72% said their organisation had faced an unexpected AI cost spike or bill in the past year, while 33% said this had happened more than once.

Visibility into the causes of those increases also appears limited. If AI spending were to double overnight, only 20% of respondents said they could identify the reason within hours.

Respondents estimated that 26% of total AI spending is wasted. Among organisations spending USD $1 million a month on AI, representing one in five respondents, that would equal USD $260,000 a month without a measurable return.

Cost ownership

The report links those pressures to day-to-day engineering work. Less than half of respondents, 45%, said they understand the cost of the AI features they build, while 56% said forecasting AI spending is based on guesswork rather than data.

Incentives also appear to play a role. Some 57% of engineers said their organisation actively encourages "tokenmaxxing," which the report described as maximising AI usage regardless of tangible value.

Most respondents said their companies already have formal controls on paper, but few have the information needed to apply them. While 73% said their organisation has AI cost policies in place, only 13% said they have basic visibility into AI spending.

The findings also suggest many companies are struggling to link spending to business outcomes. Only 26% of respondents said their organisation has a robust method for measuring the business value of AI spend.

Patrick Brogan, Director, FinOps Advisory at Harness, said the pattern resembles the early years of cloud spending, but is moving faster.

"AI spend has moved from a line item that occasionally surprises people to a budget category that regularly does," said Patrick Brogan, Director, FinOps Advisory at Harness. "The patterns we're seeing, invoice shock, ownership confusion, governance gaps, are the same ones the industry saw with cloud a decade ago, just compressed into a fraction of the time. Ownership is really the crux of it. This is fundamentally an organizational challenge. Getting teams to build with cost in mind from the start, as a design principle rather than an afterthought, is the harder and more important shift."

Engineering impact

The findings point to a disconnect between finance teams setting policy and engineering teams making day-to-day decisions on model use, prompt design and deployment. In practice, businesses may be trying to govern AI costs without real-time data at the point where those costs are created.

Harish Doddala, VP Product, Cloud & AI Cost Management at Harness, said customer discussions have shifted over the past year.

"A year ago, our customer conversations were about cloud cost attribution, commitment coverage, and rightsizing," said Harish Doddala, VP Product, Cloud & AI Cost Management at Harness. "Today those conversations are increasingly interrupted by a more urgent one: why did my AI bill do that, and how do I make sure it doesn't happen again? We're also seeing organizations write policy well ahead of building the visibility to actually enforce it, which just pushes the same problem downstream. We're seeing this pattern the same way whether we're talking to a 200-person startup or a Fortune 500 company. Cost visibility has to be part of the infrastructure from day one. Waiting until after the damage is done means teams are always reacting instead of preventing."

The report outlines a sequence it says is common among organisations with stronger control over AI costs: assigning a single accountable owner, building a unified view of spending across infrastructure, software and models, then feeding cost data into engineering workflows and tying expenditure to business outcomes.

The survey adds to a broader debate over whether corporate AI adoption is being matched by financial discipline. As more companies move AI use beyond pilots and isolated teams, the results suggest cost governance is lagging behind operational deployment, with only 13% reporting basic visibility into AI spend.