On October 19, 2026, Gartner announced that only 13% of CIOs (chief information officers) say their enterprise has realized "significant or transformational value" from AI. Companies are adopting AI rapidly, but few executives report seeing major business results. Gartner highlighted whether organizations have exit criteria: standards for deciding whether to move a pilot into production, and whether to keep a live initiative running. The findings point to the importance of not just increasing the number of AI deployments, but also setting appropriate conditions for continuing investment and revisiting them as needed.
13% of CIOs see AI value, and only 1 in 5 initiatives succeeds
The 2027 CIO and Technology Executive Agenda Survey was conducted from May to July 2026 among 2,002 CIOs and technology executives. According to Gartner's announcement, 13% of CIOs overall said their enterprise had realized significant or transformational value from AI. This figure reflects executives' assessment of value; it is not an aggregate of corporate profits or investment returns. Nor does it mean that the remaining 87% of companies are losing money on AI investments.
Gartner explains that while AI use cases have multiplied over the past two years, the success rate of individual initiatives remains at roughly 1 in 5. That rate concerns the success or failure of individual initiatives, so both what is being evaluated and the underlying population differ from the 13% executive perception of value. Even if individual AI projects deliver results, they do not necessarily add up to significant business value for the enterprise as a whole.
The published materials also do not provide a common monetary standard for judging value, the questionnaire itself, or a detailed breakdown by industry. That limits how directly a company can compare its own AI return on investment with the 13% figure. What stands out in this survey is which investment-management practices are shared by CIOs who report major results.
Exit criteria for pilots and production shape AI investment outcomes
Gartner defines "Vanguard" organizations as those whose CIOs report strong business outcomes and high AI value as perceived by the CEO. The survey found an association between being a Vanguard and having exit criteria for AI initiatives. 31% of CIOs have criteria for stopping pilots from moving into production, and 17% have criteria for ending live initiatives that deliver low value. Organizations with each type of criteria were 2.3 times and 2.6 times more likely, respectively, to be Vanguards.
| Decision stage | What the exit criteria cover | Share of CIOs with criteria | Likelihood of being a Vanguard |
|---|---|---|---|
| Pilot | Initiatives that should not move to production | 31% | 2.3x |
| Production | Live initiatives with low value that should be shut down | 17% | 2.6x |
Exit criteria at the pilot stage are meant to prevent additional investment in initiatives that fail to deliver expected results. Exit criteria in production, by contrast, are used to decide whether a system that is already running should continue. In other words, confirming some results in a pilot does not guarantee that further investment is warranted. Organizations need a mechanism to keep evaluating costs and results after launch and to decide on withdrawal when necessary. That only 17% of CIOs have production exit criteria suggests there is room to improve post-deployment investment decisions.
However, the 2.3x and 2.6x figures do not mean that introducing exit criteria will improve AI results by those multiples. The survey confirmed an association between having exit criteria and being a Vanguard, not a causal relationship. The absolute shares of each comparison group were also not disclosed, so it is not possible to calculate the probability that a given company would become a Vanguard by adopting such criteria.
Alicia Mullery, VP Analyst at Gartner, points out that "volume is not value." She stressed the basics of investment management: rather than treating a growing number of AI use cases as an achievement, organizations should define the expected business value before deployment and set exit conditions. In terms of executive decision-making, what matters is not only how many AI initiatives have been tried, but whether leaders can explain what conditions must be met to approve the next investment and at what stage to withdraw.
Pressure to cut costs is rising, yet AI's total cost of ownership is expected to grow
58% of CIOs said pressure to improve costs through AI has increased. At the same time, 51% expect AI adoption to raise the total cost of ownership (TCO) of their technology solutions. TCO refers to the full cost, including operations and maintenance, rather than only the initial purchase price. Note that 58% and 51% are the shares of CIOs who selected those responses; they are not forecasts that costs will fall by 58% or rise by 51%.
Mullery notes that many organizations continue to use budgeting and governance frameworks designed for earlier technology investments such as cloud and SaaS. Traditional technology costs are relatively predictable, whereas AI costs fluctuate with runtime processing, making it hard to forecast exact amounts in advance. As a result, CIOs are expected not just to approve upfront purchase costs but to manage the cost-effectiveness of AI throughout its operating life.
Given this cost structure, even initiatives whose value was confirmed in a pilot need their profitability reassessed once in production. Approving an investment based on pre-deployment estimates is a different matter from judging whether value continues to be generated in line with actual operating costs. The survey also does not reveal how much the 58% who reported rising cost pressure overlap with the 51% who expect higher TCO.
IT headcount expected to grow, and roles for humans and AI agents to be redesigned
40% of respondents expect to increase IT headcount in 2027. Cybersecurity and data analytics were cited as areas where hiring is especially likely. Gartner explains that as AI lowers the cost and effort of producing technology, overall demand for technology-related work will actually rise. However, the 40% figure reflects expectations about future staffing plans, not confirmation that employment has actually increased.
The spread of AI is also expected to change the scope of work that people do. Mullery gave the example of AI taking on specialized tasks so that people who once handled design also take part in testing, while testers participate in design. As such changes progress, job categories will shrink in number while the responsibilities of each role broaden. Focusing only on the possibility of AI-driven headcount reduction fails to capture this shift in job structure.
Yet only 4% of CIOs said they have a clear plan to redesign roles on the premise that humans and AI agents will work together. Organizations with such plans were 2.4 times more likely to be Vanguards. This figure also does not prove a causal effect of role redesign, but it suggests there is still a wide gap between drawing up AI adoption plans and revisiting human jobs and responsibilities to match.
The survey underscores once again that simply expanding AI investment does not lead to significant business value. When continuing to invest, organizations need to determine whether the expected business results are being achieved and whether it still makes sense to continue even after accounting for operating costs. Whether they can also revisit the division of roles between humans and AI agents and redesign how the whole organization works will be a key challenge in turning AI adoption into real business results.