Two-Thirds of CIOs Are Accountable for AI Systems They Can’t Control

Two-thirds of the world’s CIOs and CTOs are accountable for AI systems they don’t fully control, according to IBM’s new study of 2,000 C-level technology executives across 33 countries and 19 industries. The IBM Institute for Business Value (IBV), working with Oxford Economics, ran the survey from January to April 2026, and only 11% of respondents say they’re completely prepared for the scale of AI agent deployment their organizations anticipate in the next year.

Organizations relying on manual governance log 25% more AI incidents than those that embed oversight directly into their systems, IBM’s analysis found. They also deploy sixteen times fewer agents.

The Scale Mandate

The accountability gap didn’t build without pressure from above. Eighty percent of surveyed tech CxOs report a CEO-driven AI transformation mandate inside their organization, a push that has moved deployment timelines ahead of the governance infrastructure available to support them. Seventy percent say business teams are deploying technology faster than IT can track. The IBM CEO study published in the same research period, surveying 2,000 chief executives globally, found nearly two-thirds of those CEOs comfortable making major strategic decisions on AI-generated input, a posture that produces board-level delivery expectations for the CIOs and CTOs managing the actual infrastructure.

It is no longer just about deploying AI faster. It’s redesigning how organizations control, govern and invest in it and embedding control and visibility from the start, so they can scale with confidence.

Matt Lyteson, IBM’s chief information officer, offered that framing in the study’s release. His timeline is specific: by 2027, surveyed tech CxOs anticipate a 38% increase in deployed AI agents. Seventy-seven percent of organizations already report that AI adoption has outpaced their current governance capabilities, so that expansion arrives inside frameworks that were already behind before the additional volume was accounted for.

IBM’s study frames this as a design mismatch rather than an operational lag. Governance architectures in most enterprises were built for incremental, sequentially reviewed technology change. AI agents operating continuously and autonomously across interconnected systems are a different class of operational problem, and most current oversight structures weren’t built for them. IBM’s concurrent CEO study adds a longer horizon: those same CEOs expect nearly half of codifiable operational decisions to be handled by AI without human intervention by 2030, up from about a quarter today.

Accountability Without Visibility

Two-thirds of surveyed CIOs and CTOs carry direct accountability for outcomes in AI systems they don’t fully control. Security and compliance are the primary stated obstacle: 59% of respondents name them as the top barrier to scaling AI agents further.

The visibility problem has a concrete form. IBM’s IBV research published at the Think 2026 conference in May found only 18% of organizations maintain a current and complete AI inventory. The other 82% cannot accurately catalog what they’ve deployed, meaning senior technology executives carry accountability for systems their own organizations haven’t fully inventoried.

Decentralized deployment is the mechanism IBM’s survey describes. When business units spin up AI tools outside central IT review, the catalog of active systems falls further behind with each addition. IBM found 70% of organizations see business teams deploying faster than IT can track, the operational condition the IBV study ties directly to the accountability-without-visibility gap.

Victoria Medina, chief technology and data officer at Allianz Spain, contributed to the study’s executive perspectives: “AI has both a light side and a dark side. While most focus on the opportunities, it also introduces new vulnerabilities, and many organizations are more exposed than they realize.”

What Agent Incidents Cost

IBM asked surveyed organizations to count their AI agent incidents from the prior year, defining an incident as any unintended or harmful occurrence requiring human correction. That count runs across all 19 industries and 33 countries in the survey. The figure averaged 54 incidents per organization across IBM’s full survey population in the twelve-month period.

Seventeen percent were classified as high severity, requiring more than four hours to contain. At that rate, the average organization in IBM’s survey dealt with roughly nine high-severity AI incidents in a single year. When those incidents crossed the severity threshold, their consequences fell into three categories:

• 37% resulted in data exposure or security breaches
• 33% caused cascading system failures
• 17% triggered compliance issues

IBM’s analysis connected governance design directly to incident frequency. In organizations relying on manual governance, incident risk increases as AI adoption scales. Those that embed control into their systems experience 25% fewer incidents overall.

Afonso Eça, executive board member at Banco BPI, contributed to IBM’s executive interviews: “It’s like flying a plane at 10,000 feet, being told to climb to 12,000, replace both engines mid-flight and ensure zero turbulence.”

The Control Dividend

The Deployment and Margin Gap

IBM segmented the executive dataset by governance model and measured performance differences across three variables. The gap between embedded-control organizations and their manual-governance peers runs in the same direction across all three.

IBM reports relative ratios here rather than absolute agent counts. The 18% operating margin gap and the 4× budget efficiency differential apply across IBM’s full 19-industry survey population. IBM’s study does not isolate these outcomes to specific sectors or company sizes; both figures appear as consistent patterns across the full dataset.

The Financial Discipline Multiplier

A separate segment of IBM’s analysis looks at organizations with strong financial discipline in their AI programs. Those organizations deploy 2.4 times more AI agents than comparable peers without raising their overall AI or IT budget. They’re three times more likely to report being fully prepared for the scale of agent deployment their organizations anticipate.

IBM identifies the architectural decision behind the advantage. Organizations that kept AI workloads portable across environments and AI models replaceable, rather than locked into hard vendor dependencies, entered the current period of rapid agent scaling with operational flexibility that fixed-infrastructure organizations don’t have. IBM’s study tracks this as a design choice made earlier in the deployment cycle, not a characteristic of larger or better-funded organizations.

Boris Alexandre, head of the ARP programme at Airbus, contributed to the study and described the principle his organization builds around: “We design modular architectures so components can evolve as technology advances, without breaking the overall system. That approach allows us to absorb rapid innovation while supporting products with decades-long lifecycles.”

IBM’s data on this group shows those organizations posted a 10% higher return on AI investment in 2025. The study calls the underlying choice “designing for adaptability” and identifies it as the specific governance decision separating the higher-return group from the lower-return group across the survey population.

The Budget Blind Spot

AI spending is about to consume a significantly larger share of enterprise IT budgets, and IBM found most tech CxOs haven’t built the financial management infrastructure to handle that growth. The financial dimension of the governance gap runs parallel to the incident data and the performance gap, driven by the same structural conditions the IBV study documents throughout.

IBM’s study presents the financial management gap as structurally distinct from the governance architecture finding, though both appear in the same 2,000-executive dataset. Senior technology executives are directing AI investment at a scale that will approach one-quarter of total IT spend while most of them lack the real-time financial tracking to show where that investment is going or what it’s returning. IBM found no correlation between the size of AI budget growth and the operationalization of AI financial management systems to track it.

The organizations in IBM’s survey that built financial management infrastructure for AI are the same category that shows up across the study’s performance data with better outcomes on agent volume, margins, and budget efficiency. Getting there, IBM’s analysis suggests, requires treating financial visibility as a prerequisite for scale, built into the system architecture before the spending accelerates.

IBM’s survey projects AI will consume nearly 25% of IT budgets at surveyed organizations by 2027, a 71% two-year increase arriving alongside a 38% expansion in AI agent deployments. The 77% of organizations IBM found already behind on governance will absorb both pressures at once, inside the same management structures the study documents as their current baseline.