Gartner told CIOs at its Orlando IT Symposium/Xpo on October 19, 2026, that only 13% of them report significant or transformational value from AI. The odds of any single use case succeeding still sit at one in five. The finding with the most weight for an AI budget sits further down the same release. Only 31% of CIOs have guidelines for keeping a pilot out of production, and just 17% have guidelines for abandoning low-value AI that is already running. Most companies know how to start an AI project, yet few have written down what result would end one.
Principal Basil C. Puglisi addressed the pilot side of this problem in The Real State of Enterprise AI: What the Numbers Say, What Leadership Must Do, his April 2026 research synthesis. There, Checkpoint-Based Governance (CBG) moves the governance gate from the point of scale to the point of proposal. Gartner’s data supplies the other half of that gate. Puglisi’s advice to clients is to write walk-away criteria into the same approval that funds a pilot. Every stop made against those criteria then becomes a result the board can see.
What Gartner’s CIO and CEO surveys found on October 19 and 20
The CIO findings come from the Gartner 2027 CIO and Technology Executive Agenda survey of 2,002 CIOs and technology executives, fielded from May to July 2026. Gartner calls its strongest group “vanguards,” meaning CIOs who deliver better business outcomes and also report high CEO-perceived AI value. Organizations with guidelines for keeping a pilot out of production are 2.3 times more likely to be vanguards. Those with guidelines for abandoning low-value AI in production are 2.6 times more likely. Alicia Mullery, VP Analyst at Gartner, said on stage: “Volume is not value. AI is creating a reckoning for organizations that lack fundamental disciplines, such as defining use-case value ahead of time and establishing criteria for when to walk away.”
Cost pressure sits under the stopping problem. In the same survey, 58% of CIOs report more pressure to deliver cost improvement from AI, while 51% expect AI to raise the total cost of ownership of technology. Mullery described AI costs as “probabilistic and determined at run time,” so a project that looked cheap at approval can grow more expensive with every month it keeps running.
A day later, Gartner released its 2027 CEO Survey of 203 CEOs and senior business executives, conducted in April and May 2026. Forty-four percent say they struggle to show AI-driven financial returns to stakeholders, yet 75% plan to increase technology investment. David Furlonger, Distinguished VP Analyst, told the audience that “the AI value problem will not be solved by simply adding more AI to existing ways of working.”
Google Workspace published survey data the same week that shows where this pressure lands inside a company. In Beyond AI Optimism: Closing the Ambition Gap, Google reports answers from 2,638 executives and knowledge workers in six countries. Among executives, 87% say they’re being asked to show productivity gains before workflows are redesigned for AI. Seventy-two percent of companies say their AI goals already outpace their budgets. Google sells AI inside Workspace, and Gartner sells advice on these same questions, so both sources have a commercial stake in the problem they describe. Gartner’s vanguard multiples are also correlations drawn from self-reported answers, and they do not show that writing a stopping rule causes better results.
Why walk-away criteria belong in the first funding approval
Puglisi’s April paper argues that pilots stall because nobody owns the outcome in a binding way, and because pilots run on clean data with forgiving success criteria. His remedy is a proposal gate with four questions. Who is the named production owner, and what does the production integration architecture look like? What is the data quality baseline against production data, and how will the receiving workflow be redesigned? Under CBG, that owner documents the advance criteria before the pilot begins, holds intervention authority during it, and validates the result against those criteria after it ends.
Gartner’s 31% figure shows how rarely the opposite criteria get written. An advance criterion says what result moves a pilot forward, and a walk-away criterion says what result, or what missing result by what date, ends the work. When only the first exists, a pilot that misses its target usually gets another quarter and a new name, because nobody agreed in advance that a miss meant stopping.
The April paper also cites S&P Global Market Intelligence data on abandonment. In 2025, 42% of organizations abandoned the majority of their AI initiatives before production, up from 17% the year before. Read beside Gartner’s 2026 survey, the two figures suggest that companies already stop a great deal of AI work and that most of those stops happen without criteria written in advance. The surveys cover different samples and years, so this comparison is a working reading and nothing firmer. A stop that arrives through budget fatigue still teaches a company less than a stop made against a written target. Only the second one tells the next proposal where the threshold sat.
Puglisi’s distinction between Responsible AI and AI Governance decides who makes that call. In Responsible AI, the machine checks the machine, so a usage dashboard can flag a low-adoption tool on its own. In AI Governance, a named human holds checkpoint authority to accept, modify, or reject, and ending a funded program is a rejection that a person signs.
What a stopping rule changes in the pilot-to-production measure
The April paper sets a board KPI: the percentage of last year’s pilots now running in production. Its target is 40% or above within twelve months of adopting the proposal gate. Gartner’s findings put pressure on that measure, and the tension deserves to stay visible. A company that stops a weak pilot on time earns no credit under a production-only count. Worse, a team chasing the 40% target has a reason to push a marginal pilot over the line.
The answer keeps Puglisi’s measure and adds a companion. Report the share of prior-year pilots with a recorded decision, either advanced against written criteria or stopped against them, with a target of 100%. Any pilot without a recorded decision is the pilot purgatory the April paper describes, and that count is the number a board should want to watch shrink. A stop count carries its own risk, since teams can pad it by retiring work that never mattered, so pair each stop with the spend it recovered.
What clients should write down before the next AI pilot is funded
Add a fifth question to the proposal gate: what result, measured how and by what date, ends this work? Write the answer as a number against the client’s own baseline, such as cost per resolved case or hours per closed invoice. Then set a review date no more than 90 days out. The KPI is the share of newly funded pilots with written walk-away criteria, and the target is every one of them before money moves.
For AI already in production, put each system on a quarterly review that compares run-time cost per outcome against the criteria written at launch. Gartner’s 51% figure on rising total cost of ownership is the reason this review can’t wait for the annual budget, since usage-priced AI changes cost with each month of volume. Two numbers come out of that review: systems retired or scaled back on criteria each quarter, and the dollars recovered from them.
Then give the stop a line in the board report. Executives in the Google Workspace survey say they’re asked to show gains before redesign happens. A stop made on criteria is one of the few AI results a CFO can verify on the day it happens. The decided-share measure belongs on the same page as the production rate.
What to watch as 2027 AI budgets get set
Gartner’s data places the stopping decision inside a funding cycle where CEOs plan to raise technology spending while nearly half cannot yet show returns. When budgets rise and stop rules stay rare, the AI portfolio grows by default, and each new pilot competes for people and money with older ones that nobody has formally ended.
The open problem is incentive. None of the research released this week describes a company that rewards a team for ending its own project, and a stopping rule that costs its author status won’t hold for long.
Two signals are worth tracking. Watch whether the 31% and 17% figures move in Gartner’s next CIO agenda survey, and whether CEO surveys start asking how many AI initiatives were ended on purpose. Until those numbers appear, each client can build its own version from the proposal file. Count how many funded AI pilots carry a written answer to what would end them.
Sources
- Gartner. (2026a, October 19). Gartner survey finds only 13% of CIOs say their enterprise has realized significant or transformational value from AI [Press release]. Gartner Newsroom. https://www.gartner.com/en/newsroom/press-releases/2026-10-19-gartner-survey-finds-only-13-percent-of-cios-say-their-enterprise-has-realized-significant-or-transformational-value-from-ai
- Gartner. (2026b, October 20). Gartner says 44% of CEOs struggle to demonstrate AI financial returns as technology becomes central to growth [Press release]. Gartner Newsroom. https://www.gartner.com/en/newsroom/press-releases/2026-10-20-gartner-says-44-percent-of-ceos-struggle-to-demonstrate-ai-financial-returns-as-technology-becomes-central-to-growth
- Puglisi, B. C. (2026, April 1). The real state of enterprise AI: What the numbers say, what leadership must do. basilpuglisi.com. https://basilpuglisi.com/real-state-enterprise-ai-numbers-leadership/
- Snyder, D. (2026, October 20). Bridging the gap between AI ambition and readiness: New survey data. Google Workspace Blog. https://workspace.google.com/blog/product-announcements/bridging-the-gap-between-ai-ambition-and-readiness-new-survey-data
Questions readers ask
What did Gartner’s 2027 CIO survey find about AI value?
Gartner’s survey of 2,002 CIOs and technology executives, fielded May to July 2026, found that only 13% say their enterprise has realized significant or transformational value from AI. Only 31% have guidelines for keeping a pilot out of production, and 17% have guidelines for abandoning low-value AI already in production.
What is a walk-away criterion for an AI pilot?
A walk-away criterion is a written statement, set before funding, of what result or missing result by a stated date ends an AI project. Basil C. Puglisi pairs it with the advance criterion, so a pilot owner knows in advance both what moves the work into production and what stops it.
What did Gartner’s 2027 CEO Survey find about AI returns?
Gartner surveyed 203 CEOs and senior business executives in April and May 2026. Forty-four percent say they struggle to show AI-driven financial returns to stakeholders, while 75% plan to increase technology investment and 63% name radically redesigning the operating model as a top move to speed up financial returns from AI.
How does Checkpoint-Based Governance apply to AI pilots?
Checkpoint-Based Governance (CBG) is Basil C. Puglisi’s framework for placing human authority at defined points in AI work. For a pilot, a named production owner documents the criteria before it starts, holds intervention authority while it runs, and validates the result afterward, accepting, modifying, or rejecting it against those written criteria.
What should a board track for its AI pilot portfolio?
Track the percentage of prior-year pilots now running in production alongside the share with a recorded decision, either advanced or stopped against written criteria, with a target of 100%. Pair every stop with the spend it recovered, so a team cannot inflate the count by retiring work that never mattered.
#AIg
Leave a Reply