One in five AI projects succeeds.
That was the figure a year ago. And despite another year of investment, experimentation, and enterprise adoption, the number hasn’t changed.
Yet 58% of executives now report positive ROI from AI, even as organizations are spending 18 times more to achieve it than they did just two years ago.
Those figures, presented during the opening keynote at Gartner IT Symposium/Xpo 2026, raise an interesting question:
Are enterprises getting better at AI, or are they simply doing more of it?
Gartner Distinguished VP Analyst Daryl Plummer and VP Analyst Kristin Moyer challenged technology leaders to rethink how they approach AI investments, balancing the pursuit of value against growing risks.
According to Gartner, 86% of CIOs believe risk is growing faster than value, while only 25% have successfully aligned the two. Meanwhile, 71% expect to deploy agentic AI within the next 12 months, despite just 21% having governance structures in place.
The disconnect is hard to ignore.
AI adoption continues to accelerate, but the ability to consistently generate value and manage the consequences isn’t necessarily keeping pace.
More AI, Same Results
According to Gartner, AI use cases increased more than 400% in just one year.
But each new initiative is often treated as an isolated project, rather than an opportunity to build on what’s already working.
Plummer compared the approach to measuring fitness by counting the number of gyms you’ve joined.
An impressive number of memberships doesn’t necessarily mean you’re getting in shape.
“Your task is to stop compounding volume and start compounding value.”
Darryl Plummer
That distinction becomes particularly important when considering the economics behind the AI boom.
Gartner projects AI providers will spend nearly $7 trillion on infrastructure by 2030, against approximately $3.6 trillion in AI data center revenue. With that level of investment, vendors have a considerable incentive to encourage greater AI consumption.
More agents. More models. More tokens.
But more consumption doesn’t necessarily translate into better business outcomes.
While vendors frequently promote productivity improvements of 50% or more, Gartner data shows actual gains closer to 16%.
“The biggest drag on AI momentum is using AI when you don’t need to.”
Kristin Moyer
Not every task requires an autonomous agent. Sometimes conventional automation is less expensive, more efficient, and easier to manage.
Plummer called for a shift from “token maxing” to “value maxing,” encouraging organizations to focus less on expanding AI consumption and more on scaling initiatives that have already demonstrated meaningful returns.
For CIOs, that may require some difficult decisions about which investments deserve continued funding and which are adding little beyond complexity.
Efficiency Isn’t Reinvention
Even when AI delivers productivity improvements, organizations may be overlooking opportunities to create greater value.
According to Gartner, only 31% of executives are redesigning work. Most continue applying AI to existing processes rather than reconsidering whether those processes still make sense.
Moyer offered a memorable analogy.
“It’s the AI equivalent of putting a jet engine on a horse. It’s still a horse. A fast horse, but still a horse.”
The point wasn’t that efficiency gains are insignificant. It’s that making an outdated process faster doesn’t necessarily make it better.
Gartner reports that 86% of CEOs view reimagining work as a significant opportunity for AI value.
Yet the relatively small percentage of organizations redesigning processes suggests a gap between executive ambition and operational change.
For CIOs, that presents an opportunity to move the AI conversation beyond productivity metrics and into the decisions about how the business actually operates.
The Governance Gap Nobody Can Ignore
Reimagining work is one thing. Giving AI the autonomy to carry it out introduces an entirely different set of challenges.
As AI agents gain the ability to make decisions and take action across enterprise systems, traditional approaches to oversight and accountability become harder to apply.
Gartner described this shift as moving from managing “known unknowns,” risks organizations can anticipate and establish policies around, to confronting “unknown unknowns,” where harmful behavior may not become apparent until after it occurs.
“In the AI world, we are encountering the unknown unknowns,” Moyer explained. “You can’t see the harm until after it happens.”
That raises some uncomfortable questions.
- Who owns a decision made by an AI agent?
- Who is responsible when that decision causes damage?
- And how quickly can the organization intervene?
To address these challenges, Gartner introduced its “flywheel governor,” a virtual control plane connecting accountability, technology, economics, intelligence, and people.
Rather than creating another layer of centralized bureaucracy, the framework is intended to coordinate AI decisions across the enterprise.
One recommendation stood out: a rotating leadership role between the CIO and CFO, with permanent involvement from the CISO, legal teams, and other stakeholders.
It’s a notable proposal because it places financial accountability alongside technology governance.
As organizations give AI more autonomy, decisions about risk tolerance, infrastructure costs, and business value become increasingly intertwined.
The CIO can no longer address those questions in isolation.
And perhaps that’s the larger shift underway. AI governance is becoming less about approving individual technologies and more about establishing who has the authority, visibility, and responsibility to manage what those technologies actually do.
The Wrap
The first phase of enterprise AI rewarded experimentation. Organizations invested, tested use cases, and searched for evidence that the technology could deliver meaningful returns.
Now comes the harder part.
With AI portfolios expanding and success rates largely unchanged, CIOs face pressure to determine which investments are worth scaling, where existing processes need to be reimagined, and how much autonomy their organizations are prepared to manage.
Gartner’s opening keynote framed the challenge around aligning value with risk and safety with governance.
But perhaps the more important takeaway is that AI’s next competitive advantage won’t necessarily come from having access to better technology.
It may come from making better decisions about the technology organizations already have.
More AI is inevitable. More value isn’t.
