Over the past few years, businesses have been eagerly adopting AI in all sorts of ways. Yet, there’s little proof that these hefty investments are actually delivering results.
It seems many companies are now facing an “AI plateau”—a stage where initial progress has stalled and productivity gains have flatlined despite continued spending.
AI adoption is certainly climbing. The McKinsey State of AI 2025 report reveals that 88% of organizations surveyed now apply AI in at least one business function, up from 78% the year before. Still, the financial returns remain underwhelming.
Fewer than 40% of companies say their AI spending has had any measurable effect on profits. For most, that impact accounts for less than 5% of total EBIT (earnings before interest and taxes), and only 6% of organizations report an AI-driven EBIT boost exceeding 5%.
The report also notes growing interest in agentic AI, with 62% of organizations testing AI agents. While 23% claim they’re scaling agents in at least one area, no single function has seen more than 10% of companies reach “scaled or fully scaled” status.
Hidden barriers to AI transformation
All of this points to a clear pattern: AI is being used widely, but it hasn’t yet translated into meaningful business value.
According to Chris Willis, chief design officer and futurist at Domo, companies are hitting this plateau because they’re implementing AI on a technical level without making the deeper changes needed to see real impact.
Many organizations assume that simply purchasing and rolling out AI tools will automatically lead to transformation. But Willis points to several overlooked obstacles—like rigid organizational structures and a workplace culture that doesn’t encourage experimentation.
One major issue, he explains, is the false belief that every employee is naturally innovative or will spontaneously find creative ways to use new tools—especially when they’ve never been asked to before.
“AI leaders say, ‘We’re just going to hand everyone these tools and they’ll figure it out’—but that’s pure fantasy,” Willis said. “Were they even hired to be innovators? Was the accountant brought in because they love thinking outside the box?”
In reality, most companies operate as task-driven hierarchies where people are hired and rewarded for staying within their roles—not for reinventing them, he added.
Willis also notes that many organizations adopt AI out of fear of missing out (FOMO) or pressure to fix problems, without first clearly defining what those problems actually are.
If you’re trying to transform, you have to transform more than one thing—you have to rethink how decisions are made, how work gets done, and who’s accountable. Chris WillisChief Design Officer and Futurist, Domo.
“Real transformation requires overhauling multiple systems at once—not just plugging in a new tool,” Willis emphasized.
AI leadership gap
Mike Kazmier, head of AI at Banyan Software, agrees that organizations are struggling to extract value from their AI investments—and identifies two core reasons.
First, many lack strong leadership to champion AI initiatives and treat them as true business priorities. Second, they skip critical steps in the transformation process—what Kazmier calls the “six pillars of real change.”
“Too many firms believe buying the tech alone guarantees instant results, without updating their operating models, upskilling their people, or preparing their data—which is the lifeblood of any AI system,” he said.
To break through the plateau, companies need both a clear top-down strategy and a solid governance framework that manages risk and tracks value creation, Kazmier added.
Fern Halper, founder of the AI Foundations Group, notes that AI is spreading across enterprises faster than leaders can effectively manage.
Executives feel intense pressure to act quickly on AI, she said, but must also confront gaps in organizational readiness, data quality, and governance.
The speed of change has left some leaders questioning whether they can keep pace. For instance, former Coca-Cola CEO James Quincey stepped down in March as the company pushed forward with its AI and digital transformation.
Yet many organizations remain stuck in pilot mode, caught between the excitement of AI and the hard reality that foundational work—like data infrastructure and employee training—was never completed, Halper explained.
“Companies have stalled because they jumped into user-friendly AI tools without laying the groundwork: no solid data foundation, no governance, no skills development. And now they’ve hit a wall,” she said.
Not just another tech deployment
Dan Leiva, founder of CXAmplify, believes another key reason for the plateau is that companies treat AI rollouts like any other IT project—considering the job done once the system goes live.
“After deployment, nobody owns it. But that’s when the real work begins,” Leiva said. “Data drifts, upstream systems change, and without ongoing monitoring, performance degrades. With agentic AI especially, the system doesn’t stop evolving after launch—you have to keep watching it.”
Now, organizations are reevaluating how to deploy AI effectively and avoid getting stuck.
Willis suggests returning to basics: start with a few well-defined use cases where the value is clear and the data is already clean and governed.
“Integrate AI directly into existing workflows instead of building isolated tools,” he advised. “Let humans focus on what they do best—applying judgment—while AI handles repetitive tasks. Then measure the results and see if you can replicate the success.”
He warns that many teams skip straight to prototyping without asking whether they’re solving the right problem—or if the problem even needs solving.
“With automation, people rarely ask, ‘Should we automate this?’ They just do it—and end up creating new issues or using costly tech to fix something that already worked fine,” Willis said.
He draws a parallel to the early days of personal computers. A generation ago, companies equipped every employee with a PC, promising massive productivity gains—even though most workers had no idea how to use them.
There was huge investment, hype, and FOMO from leadership, paired with widespread anxiety among staff. But it took years before productivity actually improved.
“Today, we can’t imagine working without PCs,” Willis said. “But back then, it required time—not just to learn the tech, but to develop the supporting ecosystems: new roles, new skills, new processes, and a culture open to change. The same is true for AI today.”
Jim O’Donnell is a news director for TechTarget, where he covers IT strategy and enterprise ESG.
