US companies are losing 2.4% of revenue on failed AI projects

An article from

US firms lose 2.4% of revenue on failed AI projects

Businesses can curb waste by creating clear accountability structures and making honest decisions on whether projects should continue, analysts said.

Published July 2, 2026

License

Add us on Google

Getty Images

This audio is auto-generated. Please let us know if you have feedback.

The rush to deploy AI at scale is putting pressure on tech budgets as some initiatives fail to return on enterprise investments.

U.S. organizations lose an average of 2.4% of their annual revenue on AI initiatives that fail to deliver expected value, according to an Emergn report published Wednesday. The technology and management consultancy surveyed 700 senior business leaders for the report.

Just 30% of organizations said shutting down an underperforming AI or transformation initiative is considered normal practice, while nearly half said projects are typically stopped only after significant time and money have already been spent.

“The problem isn't that companies are taking risks on AI,”  said Alex Adamopoulos, CEO of Emergn. "It's that they're funding activity and calling it progress. The fix isn't cutting faster … you make the call while you still have options, before sunk costs and internal politics make the decision for you.”

Businesses are stuck between the demand for transformation and the ability to pull the plug on projects that aren’t working.

Emergn’s research shows organizations typically keep projects alive due to sunk costs, organizational politics or limited performance visibility instead of tangible evidence of business outcomes as the bills pile up.

"Before you give an initiative runway, you should be able to say three things out loud: What it's meant to prove, what would tell you it's working, and what would tell you to stop," Adamopoulos said. "If you can't answer those, you're not investing, you're hoping."

Accountability and metrics

Organizations have struggled to place governance around their AI projects, leading to runaway sprawl. As experimentation expands, businesses often find it hard to spot failing programs early and hold an accurate view of how the work is really going.

The average organization operates more than six transformation and AI initiatives simultaneously, while 1 in 10 operates with no formal oversight or governance structure at all, according to Emergn.

Christopher Panneck, head of AI, data and tech strategy at KPMG, told CIO Dive that the decision to move away from AI projects that aren’t delivering is less to do with technology and more a “culture and governance issue.”

“Clear accountability matters,” Panneck said. “When business owners are tied to outcomes, decisions to continue or stop become more pragmatic. The goal is agile governance that can keep pace with innovation: structured enough to manage risk, but flexible enough to course-correct quickly.”

Organizations should also resist judging projects by how long they've been running.

“The key is balancing patience with discipline: AI needs iteration, but not blind faith,” he said. “As one principle, don’t measure time spent, measure evidence of progress. If a use case isn’t improving outputs or changing how people work, more time won’t fix it.”

Making the right call

Without insight into ongoing initiatives, CIOs risk continuing to fund projects that no longer justify investment, Adamopoulos said.

"Only 27% of U.S. leaders said they could give their board a complete, real-time view of every transformation and AI program on demand," he said. "If you can't see what's running, you can't decide what to keep funding so cash keeps flowing to work nobody is really watching.”

Visibility is only as good as the data itself however, and Adamopoulos said bad news are sometimes “softened” on their way to the top. More than 1 in 5 respondents said project status reports present a more optimistic picture than reality, while 1 in 6 said bad news is altered before reaching senior leadership.

The response, Adamopoulos added, should be structural, making an environment where people are comfortable acknowledging bad news at any stage. Stopping projects should also become “ordinary” rather than an admission of failure, as almost a quarter of respondents said senior leaders are reluctant to admit an AI project has failed.

Sakaar Anand, chief people officer at BMC Helix, said CIOs should avoid expecting immediate returns while also creating clear decision points throughout an AI project's lifecycle.

"Giving space does not mean indefinite patience," Anand told CIO Dive. "There should be definite metrics, with an associated time frame, to make a 'kill or scale' decision. The future belongs to organizations that can successfully combine human potential with digital capability, while ensuring leadership evolves as quickly as technology,” he said.

To achieve better returns on their AI investments, businesses need to align projects with clear outcomes, review against real evidence on a set cadence and make honest decisions around what the next step should be, Adamopoulos said.

“When that rhythm is built into how you run the portfolio, stopping isn't a verdict on a person,” said Adamopoulos. “It's just good portfolio management.”