More than four in five technology leaders have canceled at least one AI pilot project because their legacy systems created limitations that made the work impossible, according to a study released Tuesday by IT services firm GFT Technologies. The research, which surveyed 945 CIOs and CTOs, found that outdated infrastructure is preventing enterprises from scaling artificial intelligence across their organizations. The findings point to a growing tension between ambitious AI goals and the aging technology foundations that most companies still rely on to run their operations.
The survey revealed widespread concern about the security implications of running AI on outdated systems, with 93% of respondents saying they believe failing to modernize their legacy infrastructure while deploying AI could eventually trigger a companywide security incident. When it comes to modernization progress, nearly half of enterprises said they've begun updating legacy systems, while roughly one-quarter reported they've started but feel they're falling behind schedule. Only 15% of technology leaders said they're nearly or fully complete with end-to-end modernization efforts, and 1% haven't started any modernization plans at all. A separate July report from Google found that just 17% of IT leaders had confidence in their technology stack's ability to support mission-critical AI agents.
Rishi Chohan, U.S. CEO of GFT Technologies, told the publication that if a CIO's goal is achieving minor productivity and efficiency gains, bottlenecks caused by legacy systems can often be managed. "But as AI grows increasingly complex and connects to more parts of any given business, like with agentic AI, legacy systems pose significant barriers to scale, speed and security," he said. The report notes that cybersecurity risks, slow release cycles, and the high costs of maintaining legacy technology were the main drivers for modernization over the last few years, but interoperability with AI systems has emerged as a priority as IT leaders devote large parts of their budgets to AI overhauls.
The study's analysis highlights that AI and agentic systems require large amounts of computing power, wide data access, persistent IT spending, and solid governance plans. As the technology spreads across organizations, legacy infrastructure is showing signs of strain under these demands. Cybersecurity risks continue to mount as AI models become more sophisticated, and AI agents breach government sites and other AI institutions. The report explains that while interoperability challenges have recently become a top concern, the costs and complexity of maintaining outdated systems have been forcing companies to consider modernization for several years, with AI adoption now accelerating that timeline.
Chohan told the publication that CIOs don't need to modernize their entire legacy technology stack at once, but should instead define their goals for how AI will work for their business, identify the systems that are holding those goals back, and modernize accordingly. "By doing so methodically, they can build the foundation their organization needs to scale AI," he said. As AI investment continues, CIOs will need to wade through the excess of AI information to clarify their vision for how it fits within their organization, and each tool a CIO chooses to deploy will require an assessment of the technical barriers preventing them from bringing their goals to life. "Few people know more about how an organization runs than its tech leaders – how it's built, how its employees work, and how its customers experience its services," Chohan said, adding that leaders are better equipped to build solutions shaped by their business needs rather than by market noise. The decision to modernize selectively rather than comprehensively may determine whether enterprises can move fast enough to capture AI's benefits before competitors do, or whether caution leaves them managing yesterday's infrastructure while trying to deploy tomorrow's intelligence.