Two out of three chief information officers are now personally accountable for artificial intelligence systems they do not fully control, according to a new IBM study of 2,000 technology executives. Boards want proof that AI spending pays off. The people expected to supply that proof are the same ones watching governance fall behind deployment inside their own companies.
Four studies published within the past four months point the same direction. Enterprise AI has moved past the adoption race and into what several executives now call its accountability phase, and CIOs are the ones whose careers increasingly hinge on the gap between what boards expect and what technology teams can actually verify.
Alignment Overtakes Security on the CIO Priority List
For the first time, aligning IT strategy with business objectives has overtaken cybersecurity as the responsibility CIOs rank most important. That is the headline finding of Experis’ CIO Outlook 2026, drawn from 1,930 technology leaders across 12 countries, up from 1,400 leaders across nine countries a year earlier. Staffing firm Experis, part of ManpowerGroup, found 48% of respondents now rank business-IT alignment as the CIO’s top responsibility, up from 34% in last year’s survey.
Cybersecurity has not become less urgent. It still tops the list of areas earmarked for bigger 2026 budgets, and the release describing the business-IT divide reaching a new high notes that the share of leaders whose risk strategy actually matches their cybersecurity readiness slipped from 77% to 72% this year. Security lost its top rank not because it got safer, but because a new demand crowded it out.
That new demand comes with real strain attached. Nearly one-third of respondents believe their own organisations may be overinvesting in AI. And 61% say the CIO’s role remains poorly understood by the rest of the C-suite, which makes proving that alignment even harder. Meanwhile, 54% say their AI investments are already generating positive returns, a genuinely encouraging number sitting right next to the doubt.
Accountable for Systems They Cannot Fully See
The sharper edge of the story comes from IBM. Working with Oxford Economics, the company surveyed 2,000 technology executives across 33 geographies and 19 industries between January and April, and found a widening gap between who answers for AI and who can actually see what it is doing.
- Two-thirds of CIOs and CTOs say they are accountable for AI systems they do not fully control
- 77% say AI adoption is moving faster than their governance frameworks can keep pace with
- 11% believe their organisation is fully prepared for the scale of AI agents expected over the next year
- 1,661 AI agents is the average number IBM expects each organisation to be managing by 2027, up 38% from today
Behind those numbers sits a pressure point that rarely makes the slide deck: 80% of respondents report that the mandate to scale AI faster comes directly from the CEO. The push is coming from above. The visibility to execute it safely is not keeping up, and a growing AI control gap at scale is what IBM’s researchers called the result. Business units are already deploying tools IT never signed off on, at a pace central technology teams cannot fully track, which means the person who answers for the outcome is often several steps removed from the decision that created it.
Why Do Most AI Projects Still Miss Their ROI Targets?
Only 28% of AI use cases in infrastructure and operations fully meet return-on-investment expectations, while 20% fail outright, leaving roughly half stuck somewhere in between. That is the finding from Gartner’s survey of 782 infrastructure and operations leaders, fielded in November and December of last year, and it lands as a blunt counterweight to any narrative of smooth AI adoption.
Gartner’s analysts argue the split has little to do with model quality. Among leaders who landed at least one successful AI use case, the deciding factors were tying the tool into existing workflows and locking in real support from business executives, not the sophistication of the underlying model. The firm’s account of AI projects stalling before meaningful ROI frames governance and business alignment as the differentiator, not engineering.
That skepticism about AI’s payoff is not confined to enterprise infrastructure teams. Similar doubts are playing out in consumer-facing markets, where leadership at companies like Chegg, Stack Overflow and Getty is now confronting the AI driven collapse of their own business models, a reminder that unproven AI value is not just an internal IT headache.
| Study | Who Was Surveyed | Headline Number | What It Shows |
|---|---|---|---|
| Experis CIO Outlook 2026 | 1,930 tech leaders, 12 countries | 48% rank alignment as the top priority | Alignment overtook cybersecurity for the first time |
| IBM with Oxford Economics | 2,000 tech executives, 33 geographies | Two-thirds accountable for AI they do not control | Governance is not keeping pace with deployment |
| Gartner | 782 infrastructure and operations leaders | 28% of AI use cases meet ROI targets | Most AI projects still miss their own goals |
| Deloitte | Global board governance research | 21% have a mature AI agent governance model | Agentic AI adoption is outrunning board oversight |
Boards Are Catching Up, Just Not Fast Enough
Boards themselves are part of the lag. Deloitte’s research on board governance found that 31% of organisations still do not have AI on the board agenda at all, an improvement from 45% previously, but still nearly a third of boards effectively delegating the entire question downward without direct oversight.
Deloitte’s argument is that AI governance has to become an operating discipline, something boards routinely test and revise, rather than a compliance box checked once a year. The firm’s broader look at progress on AI governance with room to accelerate pushes boards toward the kind of standing subcommittee structure many already use for audit and succession planning.
Sector-specific rollouts show how slowly that discipline actually spreads. Banking software provider SBS has been embedding AI tools across a 1,500-bank AI rollout still finishing in 2027, a timeline that illustrates just how long full governance and access can take to reach even a single, heavily regulated industry.
The Playbook for Winning CIOs
Not every CIO is stuck in the gap between mandate and control. Several executives interviewed this month described a similar shift: away from scattered pilots and toward infrastructure built to support AI at enterprise scale.
Vijay Balakrishnan, chief digital and information officer at Godrej Enterprises Group, said enterprises are moving from isolated AI pilots toward enterprise-wide data foundations and digital platforms capable of supporting AI at scale. He argued AI only delivers value once it is embedded into core business processes, not run as a standalone innovation project sitting beside the real business.
Balakrishnan Narayanan, chief product and analytics officer at digital lender Fibe, said the next phase of digital lending will be defined by AI that improves underwriting, fraud detection and customer experience, rather than automation for its own sake. For financial institutions, he said, the focus is shifting toward measurable business outcomes backed by alternative data and analytics.
Kaushal Kurapati, group vice-president for Oracle Fusion Applications at Oracle, described enterprise software as entering an “AI-first” era in which customers are buying business outcomes rather than applications. He argued the shift will change how enterprises evaluate software spending altogether, with success measured through productivity gains rather than how many features get switched on.
The winners will be those that combine governance, data readiness and measurable business outcomes.
Sanchit Vir Gogia, founder and chief analyst at Greyhound Research, made that case in a recent interview, arguing enterprises have moved past experimenting with generative AI and are now focused on operationalising it responsibly, not simply deploying more models. Pulling those views together, a rough playbook emerges from this year’s research:
- Build enterprise-wide data foundations before scaling new AI pilots, rather than layering AI onto fragmented systems
- Tie AI directly to measurable outcomes, such as underwriting accuracy or fraud detection, instead of automation for its own sake
- Buy and evaluate software on the productivity it delivers, not on how many features get switched on
- Integrate AI into existing workflows and secure real executive sponsorship before scaling past a pilot
The Personal Cost of a Corporate Governance Problem
What ties these findings together is where the risk actually lands. Enterprises can spread AI spending across budget lines and business units. They cannot spread accountability the same way. Boards and CEOs have one CIO to ask when a model misfires, a data breach traces back to an ungoverned agent, or a promised productivity gain never shows up in the numbers.
That is a heavier burden than the uptime and infrastructure metrics CIOs were judged on for most of the past two decades. CEOs and boards increasingly expect a CIO to show how AI improved margins, sped up product development, lifted customer experience or opened a new revenue line, on top of keeping the lights on.
Talent shortages make the job harder still. Cybersecurity remains the most sought-after technical skill in Experis’ survey, with AI and machine learning close behind, and nearly half of technology leaders named the sheer pace of technological change as their biggest organisational challenge.
Deloitte’s research offers one more marker of how far ahead ambition is running. Roughly 74% of organisations plan to adopt agentic AI within the next two years, yet only 21% currently have a governance model mature enough to manage it. The accountability gap that IBM measured this year is not closing on its own; it is set to widen right alongside the agents CIOs are being told to deploy.
Frequently Asked Questions
How is CIO performance measured differently now compared to a few years ago?
Historically, CIOs were judged mainly on uptime, cybersecurity posture, infrastructure modernisation and digital transformation milestones. CEOs and boards increasingly expect something more specific: proof that AI has improved margins, sped up product development, lifted customer experience or created a new revenue stream, on top of the older operational metrics.
Why do most enterprise AI projects fail to deliver a return?
Gartner’s research on infrastructure and operations teams found failures cluster around auto-remediation, self-healing infrastructure and agent-led management of workflows between systems, the most technically ambitious use cases. Among leaders who hit setbacks, 38% pointed to persistent skill gaps and another 38% cited poor or limited data quality as the direct cause.
What is shadow AI, and why does it worry CIOs?
Shadow AI refers to tools business units adopt on their own, outside IT’s formal approval process. IBM’s study found 70% of technology executives say teams across their business are deploying technology faster than IT can track it, which means CIOs can be held accountable for systems their own department never formally reviewed.
Which AI use cases are delivering the best returns so far?
The clearest wins are in mature, narrower applications. Gartner found 53% of infrastructure and operations leaders reported success applying generative AI to IT service management and cloud operations specifically. Separately, Experis found 34% of leaders say automation and AI-powered tools deliver the best ROI in production today, though 41% still point to cloud computing and scalable infrastructure as their top return driver overall.
What skills are CIOs prioritising as they build out AI teams in 2026?
Cybersecurity remains the single most sought-after technical skill among technology leaders, according to Experis, with AI and machine learning expertise ranked just behind it. Nearly half of the leaders surveyed named the rapid pace of technological change itself, rather than any specific skill shortage, as their organisation’s biggest overall challenge this year.
