Two-thirds of the people running enterprise technology say they are on the hook for AI systems they cannot fully see or steer.
That finding comes from IBM's Institute for Business Value 2026 CxO Study, published June 8, 2026 and conducted with Oxford Economics. IBM surveyed 2,000 senior executives responsible for IT, technology, or AI decisions, across 33 geographies and 19 industries, between January and April 2026. The headline number is the one that should make every leader pause. Roughly 67% of surveyed CIOs and CTOs report being held responsible for AI they do not fully control, as AI moves from pilots into the actual business.
If you have felt that tension in your own company, the study is telling you something useful. You are not behind. The entire C-suite is improvising at the same time.
What did the IBM study actually find?
IBM's research describes a widening distance between how fast AI is being deployed and how fast leaders can govern it. The pattern shows up across several of the survey's numbers.
Seventy percent of respondents say teams across the business are deploying technology faster than IT can track. That is the operational reality behind the accountability problem. Work gets done, tools get adopted, and the person whose name is on the risk learns about it after the fact.
of surveyed CIOs and CTOs report being held responsible for AI they do not fully control.
Source: IBM Institute for Business Value, 2026The pressure has a clear source. Eighty percent of executives report CEO-driven AI transformation mandates. Only 11% believe they are fully ready for the scale of AI agent deployment now expected. Read those two figures together and you have the core story: leadership has ordered the climb, and most organizations are not yet equipped to make it safely.
Governance is feeling the strain too. Seventy-seven percent of executives in the study say AI adoption is outpacing their governance capabilities. And the volume is about to rise. Surveyed tech leaders anticipate a 38% increase in the number of AI agents deployed by 2027.
More agents, less readiness, slower oversight. That is the situation IBM put numbers to.
Why does the 80% mandate, 11% ready split look familiar?
Because it is the same problem LaunchReady built The 7 Levels of AI Proficiency to measure.
A mandate is a decision. Readiness is a capability. When the board says "deploy AI everywhere" and the team that has to run it is not yet able to, you do not have a technology problem. You have a proficiency shortfall, and it shows up later as incidents, surprise costs, and risk nobody signed off on.
The 7 Levels of AI Proficiency exists to make that shortfall visible before the deployment, not after. It gives a leader a way to answer a plain question: where does my team actually sit, and what skill does the next level require? A Level 4 Commander in the 7 Levels of AI Proficiency, for example, is the person who can set the context an AI system runs inside. That is exactly the skill the IBM data says most organizations are missing right now. The mandate assumes the capability is there. The 11% figure says it usually is not.
Measuring proficiency before scaling is how you make the 80% mandate survivable.
What goes wrong when AI scales without control?
The study put a number on the damage, and it is not small.
Surveyed organizations experienced an average of 54 AI agent incidents last year. IBM defines an incident as an unintended or harmful occurrence that required human correction. So this is not a theoretical risk register. These are events that already happened, 54 of them per organization, on average.
The breakdown is where it gets concrete:
- 37% of those incidents resulted in data exposure or security breaches
- 33% caused cascading system failures
- 17% triggered compliance issues
- 17% were high severity, taking four hours or more to contain
Security and compliance are also what is holding deployment back. Fifty-nine percent of executives cite them as top barriers to scaling AI. The fear and the failure point in the same direction.
Afonso Eça, an executive board member at Banco BPI, described the experience this way in the study: "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." Victoria Medina, Chief Technology and Data Officer at Allianz Spain, put the other half of it plainly. "AI has both a light side and a dark side. While most focus on the opportunities, it also introduces new vulnerabilities."
What does "ready" look like in the data?
The study does not stop at the problem. It also shows what the prepared organizations do differently, and that behavior is teachable.
Organizations that embed control directly into their AI systems experience 25% fewer incidents than those relying on manual governance after the fact. Build the guardrails into the system, and you correct fewer messes later.
That is the same operating principle behind designing the workflow before you build the agent. You define how the process should run, where a human has to approve, and what the system is allowed to touch. Then you deploy. Matt Lyteson, CIO at IBM, pointed at exactly this. "It is no longer just about deploying AI faster," he said. "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."
Readiness also has a financial face that most organizations have not built yet. Eighty-four percent of surveyed organizations have not fully operationalized AI financial management, and 85% lack full real-time visibility into AI spend. If you cannot see what AI is costing you week to week, you cannot manage the return on it. That blind spot is its own kind of risk.
A few leaders in the study described the architecture choices that keep them adaptable. Dalton Gouws, Group IT Director at VWG UK, said his team is "keeping AI models plug-and-play, ready to adapt if the landscape shifts." Boris Alexandre at Airbus Canada described designing "modular architectures so components can evolve as technology advances, without breaking the overall system." Different companies, same instinct. Build for change, not for one bet.
What should a leader do with this?
You do not need to become technical to act on this study. The decisions it points to are leadership decisions.
Start with an honest read of your own team. Before the next AI mandate reaches someone's desk, ask where your people actually sit on The 7 Levels of AI Proficiency. The 80% versus 11% split in the data is what happens when that question gets skipped. A short assessment answers it in about ten minutes and tells you what the next level requires.
Then change the order of operations. The 25% fewer-incidents finding rewards organizations that design control in from the start. So design the process first, decide where a human has to stay in the loop, and build the AI to run inside those lines. Speed without that step is what produced 54 incidents a year.
And get visibility into AI spend now, while the agent count is still climbing toward that 38% increase. Real-time cost data is the difference between managing AI as an investment and discovering the bill after the fact.
Chad Jones, CIO at Baylor Scott & White Health, described the leadership posture that holds all of this together. "My role isn't to generate every transformative idea," he said. "It's to build the foundation that allows smarter people across the organization to bring those ideas to life." That is the work. Build the foundation, then scale on top of it.
Where to start
If one number from this study stays with you, make it the 11%. Only about one in nine organizations feel ready for the AI scale their CEOs are asking for, and the prepared ones got there by building control in early rather than catching up later.
You can find out where your own team stands with The 7 Levels of AI Proficiency assessment. It takes about ten minutes and gives you a clear read on the next level your people need to reach before the next mandate arrives.
Sources
- New IBM Study Finds CIOs and CTOs Face Growing AI Control Gap as Enterprise Deployment Scales
- IBM Institute for Business Value: 2026 C-suite Study (CxO)
- IBM Study: CEOs are Reshaping C-suite Roles for the AI Era
Related reading: Level 4: The Commander (Context Engineer).
Frequently Asked Questions
Is this an independent finding or an IBM marketing study?
It is IBM's own research, produced by the IBM Institute for Business Value with Oxford Economics. The numbers come from a survey of 2,000 senior executives, so treat the figures as IBM's study found rather than as settled industry fact. The pattern it describes, AI deployment outrunning governance, lines up with what many leaders are already living.
Does being accountable for AI you do not fully control mean the technology is unsafe?
No. It means oversight has not kept pace with deployment speed. The same study shows that organizations embedding control into their systems from the start see 25% fewer incidents. The risk is a readiness problem, not a verdict on the tools.
What is the single most useful number for a board conversation?
The 80% mandate against the 11% readiness figure. It captures the whole tension in one line: leadership has ordered the deployment, and most organizations are not yet equipped to run it. That is the case for measuring capability before scaling.
Find your AI Proficiency level
The free 7 Levels assessment places you across seven stages of AI capability. Under ten minutes. Research-backed scoring.