New survey finds enterprises trapped between legacy software and AI modernization

A Google Cloud executive highlights the persistent friction between new AI systems and existing legacy infrastructure, noting that cutting-edge tools often lose effectiveness when embedded in complex IT environments. The tension is intensifying as organizations adopt agentic AI workflows, with nearly 90% of the Fortune 100 using Gemini Enterprise to build and govern AI agents. However, the same legacy systems relied upon for daily operations are now acting as a drag on AI progress, creating what the survey terms an 'innovation tax.'

The survey, conducted with Hypothesis Group among 1,250 U.S. enterprise IT and business decision-makers, defines legacy technology as software running locally rather than natively in the cloud. Findings show that 49% of respondents have had AI or other technology initiatives delayed, reduced in scope, paused, or canceled due to reliance on on-premises systems. Routine maintenance and patching of these legacy systems emerged as a primary bottleneck, diverting engineering resources away from AI modernization efforts.

Security risks are compounding the challenge, as AI-driven vulnerability discovery outpaces human teams' ability to address flaws in legacy codebases. This creates a zero-sum trade-off where maintenance of outdated systems competes directly with AI adoption. The study underscores that the operational inertia of legacy software is not just a technical hurdle but a strategic impediment to AI-driven transformation across industries.

Leaders surveyed overwhelmingly support breaking vendor lock-in, with 80% agreeing that an organization’s existing software footprint should not dictate its AI choices. Meanwhile, 78% believe providers dominant in the pre-AI era may not be best positioned to lead the AI transition. The findings point to a demand for interoperable, cloud-native architectures that enable organizations to adopt best-of-breed AI tools without being constrained by legacy dependencies.