A second executive survey has landed with the same message as the first: the people accountable for AI deployments are under heavy pressure, and many cannot show what the spending has produced. Dataiku’s Global AI Confessions Report: CIO Edition 2026, released September 24, surveyed 685 chief information officers. It follows WRITER’s survey of 1,200 C-suite executives and 1,200 employees, released in April.
Along with prior surveys from McKinsey, PwC, Deloitte, Publicis Sapient, Stanford, Korn Ferry, BCG, and Gartner, these surveys paint a picture of corporate unreadiness and a good deal of uneasiness in the executive suite over organizations’ ability to harness the promised benefits of AI.
What the CIOs Said
Dataiku commissioned The Harris Poll to run an online survey of 685 CIOs between July 9 and 29, 2026, across the U.S., U.K., France, Germany, the UAE, Japan, South Korea, and Singapore. The pressure that CIOs are feeling is among the most striking.
Eighty-eight percent say their professional reputation or career will be shaped by their success with AI. Eighty-seven percent say their CEO has told them their job security depends on AI outcomes. Seventy-six percent believe their role is at risk if the company fails to deliver measurable AI gains by the end of 2027.
The pressure runs both ways. Ninety-seven percent report at least some increase in board pressure to show measurable AI ROI, and 72% say their AI budget is likely to be cut or frozen if performance targets are not met by the end of 2026.
In the U.S. sample the numbers run higher: 88% see their role at risk, 87% expect a frozen or cut budget, and 82% regret at least one major AI vendor or platform choice made in the past 18 months. The release does not publish a margin of error or subgroup sizes, so the U.S. figures are best read as directional.
The Measurement Gap
Surveys by McKinsey, PwC, Korn Ferry, Deloitte, Publicis Sapient, and Gartner have shown how little organizational productivity and ROI is being derived from AI deployments. The Dataiku report shows how ill-prepared organizations are to measure ROI.
Seventy-six percent say their organization lacks mature ROI measurement across all or most AI initiatives. Seventy-two percent cannot consistently confirm whether all their agents deliver business outcomes. Only 28% consistently measure both operational performance and business outcomes across all agents.
Cost visibility is thinner still. Only 21% have full, near-real-time visibility into AI costs attributed by business unit, team, or use case.
Accountability is also blurry. Asked who is responsible when an agent goes wrong, CIOs split four ways: shared teams (23%), central IT (21%), the data or AI team (20%), and security, risk, and compliance (18%).
The Governance Paradox
The survey’s sharpest tension sits in two numbers. Ninety percent of CIOs say they are confident they have complete tracking of all their agents. Yet 81% say they lack complete oversight of agents created outside approved systems or formal channels, and 84% agree employees are building agents and applications faster than IT can govern them.
Both can be true, since confidence in tracking approved agents says little about the ones nobody approved. Still, a CIO who is certain of full visibility while conceding that most unapproved agents escape it is describing a blind spot, not coverage.
Other figures point the same way. Eighty-three percent lack standardized agent lifecycle management. Sixty percent lack a central AI governance layer spanning enterprise apps, tools, and IT environments. Sixty-seven percent estimate 51 or more agents are running in production, and 47% have decommissioned more than 20 this year.
Dataiku’s release headline says 81% have lost oversight of their own agents. The body text is narrower, covering agents built outside approved channels. The narrower reading is the accurate one.
How It Lines Up With WRITER
WRITER’s survey, run with Workplace Intelligence from December 2025 to January 2026, asked executives and employees in the U.S., U.K., Ireland, Benelux, France, and Germany about adoption. Its findings overlap with Dataiku’s on the human side.
Twenty-nine percent of executives reported significant ROI from generative AI and 23% from agents, while 97% said AI had been beneficial overall. Forty-eight percent said adoption had been a massive disappointment. Thirty-eight percent of CEOs reported high or crippling stress about their AI strategy, and 64% feared they could lose their job if they failed to lead the transition well.
The two surveys measure different things, so the numbers should not be stacked. WRITER asked executives what return they see. Dataiku asked CIOs whether they can measure it at all. Read together, the picture is one of leaders who are unsure the payoff is there and unable to prove it either way, while their jobs ride on the answer.
WorkplaceAI covered the WRITER findings in What Can Go Wrong With AI. The wider evidence on failure rates, including a fabricated statistic that was widely repeated, is in Rectifying the AI Failure Record.
The Wider Pattern
Set beside earlier research, the two surveys fit a consistent picture. Adoption is high and rising, but earnings impact, workforce readiness, and project survival lag well behind it.
McKinsey. The State of AI in 2026, published August 25 from an online survey of 1,719 respondents in 97 countries, found 37% attribute at least some EBIT impact to AI, essentially unchanged from last year. About 6% qualify as high performers, also unchanged.
Forty-four percent say AI is scaling across the enterprise, up from 38%, and about 20% say AI operating costs, including tokens, constrain their use of it.
PwC. Its 2026 Global CEO Survey, fielded September 30 to November 10, 2025 among 4,454 CEOs in 95 countries and territories, found that 56% have seen no significant financial benefit from AI to date. Twelve percent reported both cost and revenue benefits, and 33% reported one or the other.
Deloitte. Its State of AI in the Enterprise 2026 surveyed 3,235 director-to-C-suite leaders in 24 countries. Sixty-six percent report productivity gains today. Only 34% say they use AI to deeply transform products, processes, or business models. And while 74% hope AI will grow revenue, only 20% say it is doing so now.
Publicis Sapient. Its 2026 Global Enterprise AI Report, fielded April 29 to May 14, 2026 among 1,550 AI decision-makers at large companies in six countries, found 73% use AI regularly or across most business processes. Only 10% say it is core to how the business operates. Forty-two percent say AI is capable but their organization is not set up to capture its value.
Stanford. The 2026 AI Index from Stanford HAI, as reported by Bloomberg Law, put business use of AI at about 88% of organizations, 10 points above 2024. Use is no longer the problem. Separately, a working paper co-authored by Stanford economist Nicholas Bloom found most executives reporting no measurable productivity impact, covered in Rectifying the AI Failure Record.
Korn Ferry. Its Workforce 2026 survey of more than 16,000 professionals in 11 markets found only 52% feel adequately trained to use AI. Sixty-two percent report significantly higher workloads over the past two years, and 45% say they are too busy to deliver meaningful results. Korn Ferry published these figures in a press release without fieldwork dates.
BCG. Its Global AI at Work 2026 survey found that among employees saving time with AI, 66% report little or no guidance on what to do with it, as covered in the same earlier article.
Gartner. In June 2025, Gartner predicted that more than 40% of agentic AI projects will be discontinued by the end of 2027, citing rising costs, unclear business value, and weak risk controls. It also estimated that only about 130 of the thousands of vendors claiming agent technology offer the real thing.
A project that is discontinued delivers no productivity or return, so the forecast bears directly on the ROI question. It is still a forecast, not a measured result, and it is more than a year old.
The uneasiness the CIOs describe has a basis. Companies are using AI widely, but many are not trained for it, not measuring it, and not sure which projects will survive.
One counterpoint belongs in the picture. In McKinsey’s data, enterprise scaling is rising, 80% of respondents say AI has improved their own productivity, and 60% plan to increase AI investment. The evidence shows companies struggling to turn individual gains into company results. It does not show them giving up.
Read the Source
The Dataiku and WRITER surveys were published by vendors. Dataiku sells agent management and governance software, and its release promotes those products. WRITER sells an enterprise AI platform and said its findings argue against layoffs as an AI strategy. Harris Poll and Workplace Intelligence are independent research firms, but each vendor chose the questions and the framing.
That does not make the numbers wrong. It means the questions lean toward the problem the vendor solves, and the pressure and governance findings fit neatly into a case for buying governance tools. Neither vendor report publishes a full methodology on the page we reviewed, and respondents were not broken out by company size or industry.
What It Means for Companies Deploying AI
Set a baseline before launch. An AI initiative with no pre-deployment measure of time, cost, or error rate cannot show a return later, which is the position 76% of CIOs say they are in.
Inventory every agent, including the ones nobody approved, and name one accountable owner for each. A four-way split on who owns a failure is a governance gap, not a staffing preference.
Track AI cost by team and use case now, while it is still cheap to do. And tell the board what you can measure and what you cannot, because with 97% of CIOs reporting rising board pressure, the questions are coming either way. A practical starting point is the Agent Governance Playbook for Business Leaders.
Earlier in AI Pulse: the failure record and the incident record
What Can Go Wrong With AI
Rectifying the AI Failure Record
The White House Says Reporting AI Incidents Is “Not Optional”
Bottom Line
The evidence does not say AI has no return. It says adoption has outrun readiness: the executives responsible are frightened, short on measurement, and unsure which of their own agents are running. That is a management problem that can be fixed, and it is fixable faster than a technology problem.
The companies in the best position will be the ones that can answer two plain questions: what is our AI doing, and what is it worth?
Sources: AI Pulse · Where This Breaks · workplaceai.ai. Dataiku, Global AI Confessions Report: CIO Edition 2026, press release of September 24, 2026, based on a Harris Poll online survey of 685 CIOs conducted July 9 to 29, 2026. WRITER, AI Adoption in the Enterprise, press release of April 7, 2026, conducted with Workplace Intelligence among 2,400 knowledge workers from December 17, 2025 to January 25, 2026. Additional sources: PwC, 2026 Global CEO Survey; Deloitte, State of AI in the Enterprise 2026, as reported by Unite.AI; Publicis Sapient, 2026 Global Enterprise AI Report; McKinsey, The State of AI in 2026: On the Road to ROI, August 25, 2026; Stanford HAI 2026 AI Index as reported by Bloomberg Law; Korn Ferry Workforce 2026 Insights Report as distributed by Business Wire India, September 30, 2026; BCG Global AI at Work 2026 and the Bloom NBER paper as cited in WorkplaceAI’s earlier coverage; Gartner, June 2025 agentic AI prediction as reported by Silicon.eu. The Dataiku and WRITER surveys are vendor-published; every figure above is attributed to its original publisher and none is a WorkplaceAI study. WorkplaceAI takes no position on the vendors’ products.