Can AI fix legacy IT? The ‘economics don’t hold up,’ says former IBM strategist

Only 10% of senior technology leaders report using agentic AI as a tool for modernizing legacy IT systems, according to a survey of 2,000 senior IT decision-makers conducted by IT services specialist Kyndryl. The findings, discussed by Andy Thurai, founder of The Field CTO and a former chief strategist at IBM, suggest AI implementations have so far bypassed modernization work even as vendors pitch AI as a remedy for aging systems.

The survey describes a mixed record for decades of modernization programs built around web services, microservices, cloud, virtualization and x86-based Windows and Linux servers. About half of organizations modernizing report better IT operations, and fewer than half see innovation gains. Roughly 18% say their firms found limited or unclear value, and almost half of respondents said they are behind schedule and overrunning costs. Four in ten enterprises still run mainframes.

Kyndryl reports that cost reduction and escaping legacy systems are no longer the leading motivations to modernize, with leaders instead wanting complex technology estates that mix old and new decisions to work better as AI raises demand across IT. Agentic AI — software agents that operate autonomously — is being applied to map hidden dependencies, generate code and produce documentation, tasks the report calls traditionally time-consuming and error-prone. Fewer than one in 10 organizations are fully confident they know their system dependencies today.

Thurai said modernization increasingly resembles a chronic condition, with the finish line always five to ten years out, and argued organizations remain addicted to buying software without decommissioning the systems underneath. He said three in 10 leaders already worry AI-driven modernization will produce wildly unpredictable compute costs, and that AI is the latest distraction from untangling technical debt. He added that today's AI-driven architectures will eventually become the legacy environments someone else has to modernize.

Why it matters: Teams weighing agentic AI for legacy migration have little adoption to learn from so far, and the survey links non-adopters to a higher likelihood of falling behind schedule while flagging unpredictable compute costs as a known concern.