IBM Research Finds Tech Leaders Struggle With Agent Sprawl
As enterprise technology leaders move artificial intelligence systems from pilots into production, two-thirds find themselves accountable for AI systems and outcomes that they don’t fully control, according to recent research by IBM.
See Also: AI Impersonation Is the New Arms Race-Is Your Workforce Ready?
IBM’s Institute for Business Value surveyed 2,000 technology executives from 33 geographies across 19 industries and found that CIOs and CTOs are struggling to navigate mandates from the top, agent sprawl and the unprecedented pace of change, all while managing a widening gap in visibility, governance and financial control. The study was conducted between January and April 2026 with Oxford Economics.
Transformation mandates to scale AI faster are coming from the CEO for 80% of respondents, and by 2027 those organizations expect to be managing an average of 1,661 AI agents, up 38% from today. Only 11% say they feel fully prepared to scale at this rate.
“For CIOs and CTOs, the challenge now is scaling AI systems that operate continuously and autonomously, often within governance models and architectures designed for a far slower, more predictable environment,” said Matt Lyteson, CIO at IBM, in a statement. “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.”
Shadow AI is a problem for most, and 70% of respondents said business units are deploying AI faster than they can track. These agents are creating new areas of risk, and organizations reported an average of 54 AI agent incidents in the past year, 17% of which were high-severity, taking more than four hours to contain. As a result, 37% said they experienced data exposure or security breaches, 33% dealt with cascading system failures, 17% triggered compliance issues and 13% experienced a reduction in stakeholder trust.
“AI has both a light side and a dark side,” Victoria Medina, chief technology and data officer at Allianz Spain, told IBM. “While most focus on the opportunities, it also introduces new vulnerabilities, and many organizations are more exposed than they realize.”
This “scaling trap,” between the pace of agent growth and IT’s ability to manage the sprawl, is common, the survey found. AI adoption is outpacing governance capabilities for 77% of respondents, and 59% said compliance concerns keep them from scaling agents further. More than two-thirds say business units bypass IT entirely to adopt AI tools.
CIOs and CTOs are in a bind. While prioritizing speed may make CEOs happy, business units can outpace governance. For those that prioritize a safe, slow deployment with more onerous review cycles, the cost of governance and security may be that the competition moves faster and gains an advantage.
“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. No one would choose to pilot that plane, but that’s exactly what companies are doing today,” Afonso Eça, executive board member at Banco BPI, told IBM.
Companies that deploy “orchestrated control” systems, which embed guardrails, telemetry, rollback mechanisms and kill switches directly into system architecture, are those that are seeing greater success with scaling agents, IBM said. These companies deploy 16 times more AI agents than those who rely on manual governance, the survey found.They also spend four times less of their budgets doing it, report 18% higher operating margins and experience 25% fewer AI-related incidents.
The pace and scale of AI deployments is accelerating, and AI spending is projected to grow from just under 15% of IT budgets in 2025 to nearly 25% by 2027 – a 71% increase in two years. At the same time, 84% of technology leaders said they haven’t fully operationalized AI financial management, and 85% lack real-time visibility into AI spend.
IBM advises technology leaders to concentrate their efforts in three areas to manage their agent sprawl: infrastructure adaptability, governance by design and portfolio discipline. Organizations doing all three report 38% higher expected revenue growth and 7% higher expected operating margins this year and already deploy 2.6 times more AI agents than peers.
The study finds that 88% of organizations are attempting or planning to move workloads to a different cloud provider, but technology leaders say only 25% of those workloads are easily portable. Cloud costs exceeded original projections by 48% on average, and 80% of tech leaders report data transfer costs higher than expected.
When it comes to infrastructure, IBM said organizations need to stop treating all workloads as equally important to move, and instead map lock-in exposure across infrastructure, data and AI models to identify two or three high-value portability moves that meaningfully improve flexibility without broad disruption.
Tech leaders should take a hard line on production standards to maintain governance, IBM said. If an AI agent or model isn’t registered, owned, observable and stoppable, it shouldn’t be deployed. CIOs should choose one high-impact function, like claims processing, customer service or coding and redesign governance from end to end.
From a financial perspective, tech teams should prioritize higher-value workloads and determine which AI initiatives have strategic value and which have operational value, and stop funding pilots without explicit lines of ownership and success metrics.
“AI isn’t just another technology disruption,” said Michael Voegele, global CDIO at Philip Morris International, in the report. “Unlike past innovations, AI accelerates its own development, almost like a self-fueling organism that continuously increases the pace of innovation.”
Click Here For The Original Source.
