The IBM Institute for Business Value study shows a clear “AI control gap” emerging as organizations move from pilots to enterprise-wide AI deployment.

According to the research:

• Two-thirds of surveyed CIOs and CTOs say they are held accountable for AI systems they do not fully control.
• 70% report that teams across the business are deploying technology faster than IT can track.
• 77% say AI adoption is already outpacing their current governance capabilities.

In practice, this means AI is being embedded into products, processes, and customer experiences across the enterprise, often outside traditional IT oversight. At the same time, CEOs are pushing hard on AI scale:

• 80% of tech leaders say they are operating under CEO-driven AI transformation mandates.
• By 2027, they expect a 38% increase in the number of AI agents deployed.
• Yet only 11% feel completely prepared for the scale of AI agent deployment expected in the next year.

This combination—rapid, decentralized AI adoption plus traditional governance models designed for slower, more predictable systems—creates the control gap. The study suggests that the way forward is to reimagine control and governance: embedding visibility, guardrails, and financial discipline directly into AI systems and architectures, rather than trying to bolt on manual oversight after the fact.

As AI agents scale, the IBM study finds that operational and security risks are becoming more visible and more frequent.

Key data points from surveyed organizations:

• They experienced an average of 54 AI agent incidents in the last year—situations where an unintended or harmful outcome required human correction.
• 17% of those incidents were classified as high severity, taking more than four hours to contain.
• 37% of reported incidents resulted in data exposure or security breaches.
• 59% of tech CxOs cite security and compliance concerns as top barriers to scaling AI agents.

The study also highlights a clear difference in outcomes based on how organizations design control:

• Organizations relying on manual governance see incident risk rise as AI adoption grows.
• Those that embed control directly into their AI systems experience 25% fewer incidents.

In other words, as AI agents operate more autonomously and continuously, traditional, manual oversight is not enough. The research points to the need to reshape governance so that security, compliance, and monitoring are built into AI architectures, workflows, and tooling from the start—rather than treated as after-the-fact checks.

The IBM study identifies several structural choices that separate organizations that scale AI effectively from those that struggle.

1. Build financial discipline around AI

• AI spend is expected to grow from just under 15% of IT budgets in 2025 to nearly 25% by 2027—a 71% increase in two years.
• Yet 84% of tech CxOs have not fully operationalized AI financial management, and 85% lack full real-time visibility into AI spend.
• Organizations with strong financial discipline around AI:
• Deploy 2.4x more AI agents without having a higher AI IT budget.
• Are 3x more likely to say they are fully prepared for AI scale.

2. Embed control into AI systems, not just processes

• Organizations that design control into their AI systems (rather than relying on manual governance):
• Deploy 16x more AI agents than those using manual governance alone.
• Experience 25% fewer incidents as AI adoption grows.
• This involves rethinking how AI is governed—automating policy enforcement, monitoring, and risk controls as part of the AI lifecycle.

3. Design for adaptability and portability

• Organizations that designed for adaptability early—keeping workloads portable and models replaceable instead of locked into hard dependencies—reported a 10% higher return on AI investment in 2025.
• Leaders are favoring modular architectures so components can evolve as technology advances, without disrupting long-lived products and services.

Taken together, the study suggests that to reimagine AI at scale, organizations should:

• Treat AI financial management as a core discipline, not an afterthought.
• Shift from manual oversight to embedded, automated control in AI platforms and tools.
• Invest in portable, plug-and-play architectures that can adapt as models, vendors, and regulations change.

This combination of financial rigor, built-in governance, and adaptable architecture helps technology leaders scale AI with more confidence and clearer accountability.