AI Legacy IT Modernization Struggles as Only 10% Deploy Agentic AI, Economics Remain Questionable

Legacy IT Modernization Faces Persistent Challenges Despite AI Promises

Modernization has dominated CIO agendas for decades, yet the finish line remains perpetually out of reach. A 1

1

of 2,000 senior IT decision-makers reveals a sobering reality: only 10% of tech chiefs currently apply agentic AI for modernization efforts. This statistic underscores how AI implementations have largely bypassed technology modernization initiatives, even as organizations invest heavily in AI models and autonomous agents.

The results paint a troubling picture of modernization outcomes. Only about half of organizations modernizing their systems report achieving better IT operations, while less than half see innovation gains. Perhaps most concerning, 18% of firms report limited or unclear value from modernization efforts. Almost half of respondents indicated their projects are behind schedule and suffering cost overruns, highlighting the persistent difficulties in executing these critical initiatives1

1

.

The Economics of AI-Driven Modernization Don't Hold Up

Andy Thurai, founder of The Field CTO and former chief strategist with IBM, offers a stark assessment of the current landscape. "IT modernization increasingly looks like a chronic condition," Thurai stated, noting that four in ten enterprises still run mainframes despite modernization being positioned as a top mandate for years. The finish line, he observes, remains perpetually 5 to 10 years away1

1

.

Thurai warns that vendors pitch AI to IT leaders as a cure-all, but "the economics don't hold up." When AI drives infrastructure sprawl, monthly bills become black boxes. The Kyndryl survey found that three in 10 leaders already worry AI-driven modernization will result in wildly unpredictable compute costs. Organizations remain addicted to buying software while legacy systems underneath almost never get decommissioned, with AI becoming the latest distraction from untangling technical debt1

1

.

Agentic AI Shows Early Promise for Mapping Dependencies

Despite economic concerns, agentic AI demonstrates early potential to address longstanding modernization challenges. Organizations using agentic AI for modernization are significantly less likely to fall behind schedule compared to those not adopting this approach. The technology helps map hidden dependencies, generate code, and create documentation—traditionally time-consuming, error-prone processes that consume months of manual labor1

1

.

Fewer than one in 10 organizations express full confidence in understanding dependencies within their technology estates today, representing a significant technical blind spot. Agentic AI operates autonomously while providing enhanced visibility on dependencies between systems, both legacy and new. Organizations that moved past bolt-on chatbots are deploying agentic frameworks as orchestration layers to map technical dependencies and generate missing documentation as they proceed1

1

.

Knowledge Debt Exceeds Technical Debt as Primary Challenge

The conversation around legacy IT systems typically centers on outdated technologies and reducing technical debt. However, the greater challenge is knowledge debt. Legacy applications contain decades of business logic, operational processes, regulatory requirements, and customer rules reflecting how organizations price products, manage risk, process transactions, and respond to exceptions2

2

.

Rather than viewing these systems purely as technology requiring replacement, enterprises should treat them as repositories of business intelligence. Critical business processes remain spread across decades-old applications, siloed workflows, and disconnected systems, creating barriers to agility, innovation, and scale. The ability to unlock and reuse this embedded knowledge will increasingly determine how effectively enterprises can adapt, innovate, and deploy AI for modernization at scale2

2

.

AI-Native Engineering Requires Understanding Before Transformation

Many transformation initiatives fail because organizations attempt to modernize legacy IT systems before fully understanding them. Applications that appear outdated often support hundreds of dependencies, undocumented business rules, and critical integrations. Replacing or refactoring them without sufficient visibility creates significant business risk2

2

.

AI changes this equation by enabling organizations to analyze codebases, map dependencies, uncover hidden business logic, and document workflows at previously impractical scales. This creates a clearer picture of what should be rebuilt, refactored, retained, or retired. The objective isn't to automate modernization blindly but to improve decision-making, reduce uncertainty, and establish a stronger foundation before transformation begins2

2

.

Shifting Focus from Cost Reduction to Business Capabilities

Cost reduction and legacy escape no longer lead modernization motivations. As AI increases demand on nearly every IT element and sovereignty considerations become more acute, leaders want their complex technology estates—a mix of old and new decisions—to work better. Cloud migration and microservices aimed to create more advanced and flexible systems, but modernization must become more than swapping out mainframes for something sleeker1

1

.

The focus is shifting toward expanding tangible business capabilities through AI-native engineering. This approach emphasizes creating an intelligent engineering lifecycle that can continuously understand, build, modernize, and improve software. Engineering intelligence connects business intent, architectural decisions, development workflows, testing outcomes, and operational insights into a continuous flow, making the entire engineering system more responsive, predictable, and adaptable2

2

.

Balancing Automation with Governance and Human Oversight

Enterprise AI adoption highlights that speed without trust creates risk. While AI can accelerate analysis, development, and testing, organizations need confidence in accuracy, security, compliance, and reliability. Enterprise software cannot rely on automation alone—it requires governance, validation, traceability, and human oversight2

2

.

Successful adoption depends on balancing automation with control through structured workflows, policy guardrails, auditability, and clear accountability. This ensures AI-generated outputs remain aligned with business objectives and regulatory requirements. Moving from experimentation to enterprise-scale deployment requires both intelligence and trust, particularly as organizations navigate the complexities of AI-driven innovation while managing unpredictable compute costs and infrastructure sprawl2

2

.