Three years ago, most chief technology officers would have told you AI was the easiest bet they’d ever made. Budgets grew, pilots multiplied, and confidence in scaling the technology sat above 80 percent. That optimism has now cooled for a third year running, and the reasons behind the drop say more about how enterprises actually work than they do about the technology itself.
A new report from Akkodis, the digital engineering arm of the Adecco Group, surveyed 500 CTOs as part of a broader study of 2,000 C-suite executives. The headline number is stark: confidence in an organization’s ability to scale AI has fallen to 48 percent in 2026, down from 62 percent in 2025 and 82 percent in 2024. That’s a 34 point drop in two years, at the exact moment AI spending is supposed to be paying off.
Why the Drop Isn’t About the Technology
It would be easy to read this as AI fatigue, a sign that the models simply aren’t good enough yet. The report suggests otherwise. When CTOs were asked what’s actually holding scaling back, the top answer was a lack of in-house AI skills, cited by 32 percent of respondents. Uncertainty about return on investment came in close behind at 31 percent, followed by insufficient internal urgency to act at 27 percent.
None of those are technical limitations. They’re organizational ones. Only 44 percent of CTOs believe their leadership teams actually understand AI well enough to make good decisions about it, and just 46 percent say their organization has established frameworks for responsible AI use. Workforce trust is arguably the weakest link of all: only 36 percent of CTOs are satisfied with how much their employees trust the AI systems they’re being asked to use.
That distinction, between deploying a tool and actually integrating it, echoes a pattern seen elsewhere in enterprise AI. We covered a similar dynamic in our look at why the vast majority of AI projects fail to deliver value, where analysts pointed to the same culprits: unclear ownership, weak partnerships, and a habit of throwing money at exploratory projects instead of building the processes to support them.
Agentic AI Is Raising the Stakes
Complicating matters further is the rise of agentic AI, systems that can plan, make decisions, and carry out multi-step tasks with minimal human oversight. Forty percent of CTOs named it the single most impactful technology trend shaping their organizations this year, more than any other category. Fifty-seven percent already use some form of AI to decide which tasks should go to a person and which should go to a machine.
That’s a meaningful shift in responsibility. Governance frameworks built for a chatbot that drafts an email look very different from what’s needed for a system that can approve a transaction or reroute a supply chain on its own. It’s little surprise that the CTOs least confident about scaling AI are also the ones who feel weakest on governance and accountability, since agentic systems raise the cost of getting either one wrong.
Jo Debecker, President and CEO of Akkodis, framed the shift plainly: "What we’re seeing now is not a slowdown in AI adoption, but a moment of realism. Organizations are moving beyond experimentation and encountering the reality of scaling AI across complex environments. The challenge is no longer deploying AI, it’s integrating it into how work gets done. The companies making progress are those redesigning their operating models, aligning technology, human expertise and governance to deliver consistent results."
The Legacy Systems Problem Nobody Wants to Talk About
Part of what makes integration so hard is that most large organizations aren’t building on a blank slate. They’re layering agentic AI on top of decades-old enterprise systems, patchwork data pipelines, and processes that were never designed with automation in mind. That mismatch is a bigger drag on AI scaling than most roadmaps admit, and it’s a big part of why technical debt keeps surfacing as a hidden blocker in survey after survey.
Organizations that have made real headway on this front tend to treat legacy modernization as a prerequisite for AI, not an afterthought. Our breakdown of practical ways to use AI to modernize legacy systems covers exactly this kind of groundwork, and it’s a useful checklist for any CTO staring at a scaling plan that assumes infrastructure nobody actually has.
A Genuine Strategic Shift, Even Amid the Caution
There’s a more encouraging thread buried in the same data. For the first time since Akkodis began running this survey, CTOs cited innovation, not efficiency, as the primary driver of digital investment. That’s a real change in posture. Cost-cutting justifications got AI budgets approved in the early days; now boards are being asked to fund AI because it opens new revenue lines and business models, not just because it trims headcount.
That mirrors a broader pattern showing up in enterprise research this year: the companies pulling ahead with AI aren’t necessarily the ones with the biggest budgets or the flashiest models. They’re the ones using AI to chase growth rather than just shaving costs, and letting it make more decisions without a human double-checking every step. We explored this gap in detail in our piece on why AI’s biggest productivity gains are still ahead of us, which found that a small share of companies are already capturing a disproportionate amount of AI’s economic value while everyone else is still stuck running pilots.
What This Actually Means for Enterprise AI in 2026
Put together, the picture isn’t one of AI losing momentum. Investment keeps climbing and agentic AI keeps gaining ground. What’s falling is the assumption that deploying the technology is the hard part. It never really was. The hard part, as this report makes clear, is redesigning how an organization actually operates so that AI fits into the work rather than sitting on top of it.
For CTOs feeling less confident than they did a year or two ago, that’s not necessarily a bad sign. A dose of realism after two years of unchecked optimism tends to produce better decisions than blind faith ever did. The organizations that treat this moment as a chance to fix governance, close the skills gap, and rebuild trust with their own workforce are the ones likely to be the confident 48 percent again next year, and maybe even higher.
