AI spending may slow down as ROI remains elusive

Most enterprises have yet to see their AI investments make a positive impact on their bottom lines, which could lead many to rethink AI budgets over the next year, IT analyst firm Forrester predicts.

Many enterprises will delay a quarter of their planned AI spending until 2027 as they struggle to see a return on investment, Forrester projects. Just 15% of AI decision-makers surveyed by the analyst firm report AI-related earnings increases for their organizations in the past year. Less than a third can link AI to income hikes.

“The disconnect between the inflated promises of AI vendors and the value created for enterprises will force market correction,” Forrester’s analysts write. “Savvy buyers should capitalize on this supply side frailty by manipulating the levers of AI cost while refocusing investment on top- and bottom-line impact.”

Many CIOs who talk to Brian Hopkins, vice president of emerging technology at Forrester, have recently voiced concerns about a lack of ROI in their AI projects. While many are observing efficiency gains, it’s difficult to translate a 15-minute savings when an employee writes an email to an improvement in the bottom line, he says.

“Because firms are having a hard time moving from individual task kind of efficiency into more process efficiency, they’re not seeing any earnings at the top line,” he says. “Can we say they’re not getting the benefits, or are they just not measuring the full value that they’re getting because they’re struggling to connect what they’re doing?”

Hopkins expects AI spending delays during 2026 to be common, with about half of the organizations in Forrester’s financial services- and healthcare-heavy client base putting off planned outlays.

CIOs, pushed by company executives to launch AI projects, have questioned spending on big AI platform deployments, Hopkins says.

In one recent conversation with Hopkins, a company CIO worried about the purchase of an AI platform. “The CIO is like, ‘Is the value going to happen, or do we just have to wait a little while?’” he says. “There’s a lot of these big platform investments, and what’s going to happen when the value doesn’t materialize is that spending is going to get cut.”

Some don’t see a bubble

Some AI experts agree with Forrester that an AI market correction is on the way. Microsoft founder Bill Gates recently talked about the existence of an AI bubble, and industry observers have noted that some AI excitement is dimming. Many don’t see an AI bubble that will burst in the near future, but it’s deflating a bit.

Still others don’t see much of a slowdown in the near term.

Hightouch, provider of an AI and data platform for marketing, doesn’t see indications of a spending slowdown, says Brian Kotlyar, CMO there. Instead, organizations are getting pickier about what they’re spending their AI dollars on, he says.

“What we do see is a rapid increase in sophistication in understanding AI and how to use it,” he adds. “The budgets remain allocated and there remains urgency to spend money to unlock the benefits of AI, but the buying criteria are more mature and stringent than the initial rush of experimentation.”

Still, Kotlyar doesn’t rule out a market correction at some point. He notes the tulip buying craze of the 1630s, when bulb prices reached incredible levels.

“Two things are happening at the same time,” he says. “The underlying utility of AI for labor augmentation and replacement is clear and indisputable. At the same time, markets can get exuberant — there was a time that you could trade a single tulip bulb for a whole house in Amsterdam, after all.”

Even with a possible slowdown at some point, there’s a clear AI trend, he adds. “AI is very good today with zero further advancement at a lot of very valuable work,” Kotlyar says. “It’s not going away.”

Slowing down

Other observers see the potential for an AI spending slowdown, as organizations move away from experimental AI projects and refocus on revenue-generating solutions.

Some organizations are not achieving the accuracy they need from AI tools, and others are not finding their data to be easily accessible or properly structured, says Sam Ferrise, CTO of IT consulting firm Trinetix.

“Many organizations are realizing that their expectations for AI accuracy and performance don’t always align with the level of investment they’re willing — or able — to make,” he says. “The key is calibrating expectations relative to both the investment and the use case.”

In other cases, enterprises deploying AI are running into privacy or security problems, he adds. “Many teams successfully prove a use case with clear ROI, only to realize later that they must harden the solution before it can safely move into production,” Ferrise says. “When that alignment isn’t there, it’s natural for organizations to pause or delay spending until they can justify the value.”

The prospect of a bubble bursting may be an overly dramatic scenario, although not impossible, he adds. It’s been easy for organizations to overlook intangible costs such as training, compliance, and governance.

“What seems more plausible is a correction or reset,” Ferrise says. “Companies will likely narrow their focus to use cases that require lower investment but deliver higher perceived value.”

An emphasis on AI outcomes

Srikrishnan Ganesan, co-founder and CEO of agent-based professional services automation vendor Rocketlane, agrees that experimental AI spending is likely to go down in 2026, as companies look for AI projects with proven results.

“We’ve been in the vision-selling era for too long and need to move to the outcomes era,” he says. “There will be corrections in the market expectations because of a lot of overlapping ambitions today among companies claiming to drive huge ROI with AI.”

Still, there will be many organizations that try to move toward “radical efficiency” with AI, Ganesan adds. “The overall AI spend should still increase as production AI deployment reaches more departments in the enterprise,” he says.

AI users are moving from proof-of-concepts to production readiness, leading to slower headline spending and smarter allocations, adds Dan Zimmerman, chief product and technology officer at TreviPay, a global B2B payments platform.

“Many organizations are realizing their early AI pilots were heavy on experimentation and light on measurable ROI, requiring a necessary recalibration to refocus on outcomes,” he says. “The period of hype and excitement around AI, where everyone was experimenting, making bold promises, and rushing to launch pilots, is ending, and we’re entering a more disciplined, technical, and practical phase.”

Organizations that invested in quick demos may pull back, but those integrating AI into core workflows such as customer service and credit decision-making, are seeing sustained ROI, Zimmerman adds.

“I can say with confidence AI is improving productivity and making knowledge more accessible across the business,” he says.