The Number That Should Worry Every CIO
Gartner's 2027 CIO and Technology Executive Agenda survey, fielded between May and July 2026 across 2,002 CIOs and technology executives, put a hard number on a frustration almost every tech leader already felt: only 13 percent of enterprises have realized significant or transformational value from their AI investments. That figure is the headline finding from the single largest annual survey of CIOs Gartner runs, and it lands after three straight years of rising budgets, vendor hype cycles, and board-level mandates to do something with AI. The gap between spend and proof has become the defining tension of enterprise technology strategy heading into 2027.
Gartner analyst Alicia Mullery, speaking at the Gartner IT Symposium/Xpo in Orlando where the findings were presented, offered a blunt diagnosis: "Volume is not value." Enterprises have shipped pilots, chatbots, and copilots by the dozen, but Gartner's survey finds that any individual AI use case still carries roughly a one-in-five chance of succeeding. For a CIO reporting to a board that assumes adoption automatically translates into returns, that ratio is the uncomfortable fact that needs saying out loud before the next budget cycle, not after it.
Volume Without Value
The instinct inside most IT organizations has been to scale AI experimentation as fast as possible, on the theory that more pilots mean more chances at a breakthrough. Gartner's data suggests that instinct runs backward. Launching more pilots without a disciplined way to kill the ones that are not working simply multiplies sunk cost and enlarges the eventual reckoning. Mullery's framing is aimed squarely at CIOs who have been measuring AI progress by activity, deployment counts, licenses purchased, use cases launched, rather than by outcomes delivered against a defined business case.
This matters because related coverage from Campus Technology, reporting on the same survey cycle, found that 75 percent of CEOs plan to keep increasing technology budgets even as measured AI returns lag. That combination, rising spend paired with flat proof, is exactly the setup that turns an innovation program into a credibility problem for the CIO. Boards eventually ask for the return, and a portfolio built on volume rather than governed experimentation will have no clean answer ready when they do.
The Governance Gap Nobody Budgeted For
The most actionable finding in the survey is not about AI capability at all. It is about governance infrastructure. Only 31 percent of CIOs report having formal guidelines for when a pilot should not advance to production, and just 17 percent have criteria for abandoning a low-value AI initiative that has already made it into production. Most enterprises have built elaborate pipelines for starting AI projects and almost nothing for stopping them, which is precisely how low-value pilots accumulate into the kind of bloated, unaccountable portfolio that eventually draws board scrutiny.
Gartner calls the CIOs who have solved this the vanguards: leaders delivering better business outcomes whose CEOs perceive high AI value. Vanguards are 2.3 times more likely than their peers to have explicit stop criteria for pilots, and 2.6 times more likely to have criteria for retiring underperforming production initiatives. The lesson for everyone else is specific and operational: write the kill criteria before the pilot starts, not after it has already consumed two budget cycles and a chunk of political capital.
Cost Pressure Is Rising Faster Than Proof
Fifty eight percent of CIOs report increased pressure to deliver AI cost improvements, and 51 percent expect AI to increase the total cost of ownership of their technology stack rather than reduce it. That second figure deserves attention from every CFO who has budgeted AI as a cost-saving initiative. Mullery noted that AI costs behave differently from the technology spend finance teams are used to managing: "Unlike traditional technology costs, which are largely predictable, AI costs are probabilistic and determined at run time," meaning usage-based inference pricing can swing with adoption, prompt complexity, and model choice in ways a fixed software license never did.
That probabilistic cost structure is itself an argument for the governance discipline Gartner describes. A CIO who cannot kill an underperforming AI initiative is also, by definition, unable to cap its runaway inference costs. Cost governance and value governance are the same muscle, used twice. CIOs treating AI procurement like a traditional software purchase, fixed budget, fixed scope, annual renewal, will keep getting surprised by both the bill and the board's questions about why the bill does not match the promised value.
The Workforce Redesign Nobody Is Doing
Perhaps the starkest number in the survey is this: only 4 percent of CIOs have concrete plans to redesign roles for a human-agent workforce, even though vanguards are 2.4 times more likely to have such plans than their peers. Forty percent of respondents do expect to increase IT headcount in 2027, concentrated in cybersecurity and data and analytics, which signals organizations are still staffing up to support AI rather than restructuring work around it. That distinction, hiring more people to manage AI versus redesigning jobs so people and agents work together, marks the gap between automation theater and actual transformation.
This finding should reframe how CIOs talk about AI maturity internally. A chatbot deployment count or a model fine-tuning roadmap is weak evidence of transformation if the underlying org chart, approval workflows, and job descriptions look exactly as they did in 2023. Gartner's vanguards are distinguished less by which models they use and more by whether they have done the unglamorous work of rewriting who does what when an agent can execute a task a human used to own.
What This Means for Your Roadmap
For a CIO or CTO heading into 2027 planning season, Gartner's findings translate into three concrete actions. First, every AI initiative entering a pilot needs written, board-visible criteria for what failure looks like and when the project gets killed, set before it launches, not after a year of sunk cost. Second, cost governance for AI needs to be modeled as a variable, usage-driven expense from day one, not bolted on after the first surprising invoice. Third, workforce redesign planning needs its own line item and owner, separate from the AI tooling roadmap, because headcount growth without role redesign is not transformation.
None of this requires slowing AI adoption down. It requires treating AI governance as a prerequisite for scale, the discipline that separates Gartner's vanguards from the 87 percent of enterprises still waiting for AI to pay off. The CIOs who build stop criteria, cost models, and workforce plans now will be the ones with a credible answer when their board asks the question every board is about to ask: where is the return.
