AI’s ROI Divide: Why Enterprises Are Still Struggling to Cash In

For all the hype around artificial intelligence, one truth is becoming increasingly clear: the organizations that reap the biggest rewards aren’t the ones throwing money at AI, but the ones that have already mastered the messy, unglamorous world of information management.

That’s the conclusion of new global research commissioned by OpenText, which surveyed more than 500 CIOs and senior IT leaders across industries. The study highlights a widening gap between seasoned AI adopters and those still in their early phases. Nearly 70% of mature AI users reported being highly satisfied with their return on AI investment (ROAI). Among newer adopters, that number plummets to just 42%.

A financial services CIO summed up the problem bluntly: “Our business wants more AI and our security and IT are panicking because nobody knows how to secure it and run it in a well-governed way.”

When Governance Becomes the Bottleneck

The survey suggests that AI value isn’t about how much an enterprise spends, but whether it has the right guardrails in place. Security and compliance concerns remain the number one information management hurdle, with 44% of respondents naming it as their biggest barrier. Less than half rated their governance programs as working “very well.”

That governance shortfall has real consequences. A healthcare VP quoted in the research stressed that, “AI success depends on better metadata management, stronger governance, and classification.”

It’s not just about preventing breaches—it’s about ensuring data is structured and trusted enough to fuel AI systems responsibly. Without that, AI projects tend to stall or produce shallow results.

The Payoff of Maturity

The study found that enterprises spending more on AI—an average of $5.4 million annually on generative AI tools, infrastructure, and talent—are also the ones that invest in robust data governance. Mature adopters measure success by problem-solving impact and risk reduction, while newer entrants tend to measure efficiency alone.

Multi-agent AI, in particular, is seen as transformative by those further along the curve, enabling organizations to tackle more complex business problems rather than just shaving minutes off repetitive workflows.

“AI’s value does not appear overnight. Real returns come when enterprises move beyond surface-level automation to solve real problems. That takes secure, well-governed information. Without that foundation, ROAI is almost impossible to achieve,” said Savinay Berry, Chief Product Officer and Chief Technology Officer at OpenText. “We give organizations that foundation by simplifying information complexity, strengthening governance, and ensuring information is ready for AI.”

The Trust Factor

Interestingly, satisfaction with AI results correlates closely with whether leaders trust their tools to safeguard information. Companies confident in their governance posture were far more likely to say their AI investments were paying off.

And while maturity levels differ, both early and advanced adopters ranked “deep expertise in secure information management” as the number one quality they look for in a partner—underscoring how fundamental data trust has become to AI strategy.

The Bottom Line

The OpenText research reinforces what industry veterans have been saying quietly for years: AI doesn’t fail because the models aren’t powerful enough. It fails because the data they depend on isn’t ready.

In other words, if CIOs want to close the satisfaction gap, they need to worry less about the next flashy generative AI demo and more about the unsexy plumbing underneath: governance, classification, and secure data foundations.

Otherwise, that ambitious AI budget will keep returning underwhelming results—and enterprises will remain stuck in the shallow end of the AI pool.