What ROI? AI misfires spur CEOs to rethink adoption

Nearly three years after the release of ChatGPT kicked off a new revolution, consistent AI ROI remains elusive.

Just 25% of AI initiatives in recent years have lived up to ROI expectations, according to CEOs surveyed by the IBM Institute for Business Value. Meanwhile, organizations have achieved enterprise-wide rollouts with only 16% of AI projects.

Part of the problem has been that organizations don’t know what they’re getting into. Nearly two thirds of CEOs acknowledge that the fear of missing out drives investment in new technologies before they have a clear understanding of its value. Often this results in rushing an AI project for ROI or launching one just for show — neither a recipe for success.

IT experts who’ve witnessed the AI gold rush of the past two and a half years aren’t surprised at the lackluster performance of AI projects.

“These survey results are very much in line with what we see across the industry,” says Ivan Navodnyy, chief product officer of fintech solutions provider B2BROKER. “Forget about delivering measurable ROIs; just a small fraction of AI projects even makes it to production stage in the first place.”

Many organizations have jumped into AI because it’s a “trendy” technology, but leaders haven’t figured out what problem they’re trying to solve, Navodnyy adds.

“They want their brand to be seen as AI-first but often apply it to non-urgent problems — like experimenting with content generation, for example, instead of identifying a core business issue to fix,” he says. “It’s not about simply jumping on the AI train. It’s about being sure that you’re riding it in the right direction and that you even need it in the first place.”

Not enough expertise

Many IT and business leaders have rushed into AI adoption without considering internal expertise or the need to sell the technology to internal users, prompted in part by a fear of missing out, Navodnyy contends.

“When leaders feel the pressure to move fast just to stay competitive, they can sometimes skip critical steps, prioritizing deployment speed over product quality,” he explains. “That’s a very reckless approach, and an easy way to end up with wasted resources and damaged reputation.”

The speed of the AI revolution has contributed to CEOs jumping into the technology before their organizations were ready, says Neil Dhar, global managing partner at IBM Consulting.

AI’s pace of change has created concerns among CEOs about anticipating business needs, he notes. Forecasting accuracy was the top concern for CEOs in the 2025 IBM survey, while product innovation was the top concern in 2024, and productivity and profitability were top issues for CEOs in 2023.

“CEOs have to constantly look around the corner and look at what disruption would look like in their respective businesses, in their industries,” Dhar adds. “As they do that, they need to constantly be aware of macro trends, and they have to react.”

Still, many organizations moved forward before they were ready to succeed with AI, Dhar adds. For example, many companies launched initiatives without having cleaned up their internal data for use with AI, he says.

“Because AI came on us so fast, there were a lot of folks who started skunk-work exercises for certain things that maybe impact a very small part of an organization versus looking at the wider picture and saying, ‘Does this have enterprise value?’” he says.

Evaluate the risks and benefits

In addition to not laying a strong foundation for AI success, many organizations have failed to project the time and investment needed to achieve ROI with their AI projects, says Nagmani Lnu, director of quality engineering at financial services firm SWBC.

“Before introducing AI, we must ensure that we have the right ecosystem to train the model effectively to ensure a perfect result,” he says. “Mistakes will be costly. We want to focus on the low-hanging, least risky area.”

Despite past mistakes, organizations still need to embrace the technology, IBM Consulting’s Dhar says. “This technology is going to continue to evolve, and it’s going to have profound impact on every industry, every function,” he adds. “If you don’t really figure out what your gen AI story is, then you’re going to be left behind.”

However, CEOs may be starting to rethink the way they adopt AI, according to the survey. Only 37% of the CEOs said it’s better to be “fast and wrong” than “right and slow” when adopting new technologies — a shift from gen AI’s early days.

About two-thirds of the CEOs say they’re now leaning into AI use cases based on ROI, signaling a change from the past couple of years. CIOs are also changing their tune on AI, pivoting to practical applications over experimentation of late.

Beware underestimating the effort needed 

Organizations need to look before they leap, SWBC’s Lnu adds. “Slow and steady wins the race,” he says. “Understand risk before moving and evaluate whether ROI is compelling and goes beyond the risk.”

Most companies have significantly underestimated the effort it takes to move from a promising AI prototype to a production-level system delivering value at scale, says Zulifkar Ramzan, CTO of Point Wild, parent company of several cybersecurity providers.

Many AI projects fail not because the technology is flawed, but because the organization lacks the internal skills needed to make it work, he says.

“AI is a powerful tool, but a fool with a tool is still a fool,” Ramzan adds. “Writing the core AI code might take a single engineer a few days to a week — but getting that capability into production can take months and a village.”