On October 8, 2026, IT services specialist Kyndryl published data showing that only 10% of surveyed senior IT decision-makers deploy agentic AI legacy IT modernization tools to tackle aging system architectures ZDNet. The finding highlights a sharp division between vendor pitches promoting artificial intelligence as a universal fix for infrastructure debt and the financial reality inside enterprise data centers.
The Economics of Legacy Overhauls
For decades, enterprise technology budgets absorbed continuous streams of modernization projects. Organizations ported applications from traditional mainframes to on-premises architectures and cloud environments through web services, microservices, virtualization, and x86 hardware migrations. Despite these repeated investments, outcomes remain mixed. Kyndryl researchers report that roughly half of enterprises undertaking modernization achieve improved IT operations, while less than half record measurable innovation gains.
Financial and schedule metrics paint a starker picture. Nearly one in five organizations report limited or unclear value from past modernization initiatives. Almost half of the respondents surveyed note that their projects suffer from severe cost overruns and missed deadlines. These operational friction points challenge the traditional assumption that swapping out foundational systems yields immediate financial returns or structural agility. Enterprises face steep compute and engineering bills whenever proprietary business logic requires translation into modern languages, stretching payback periods beyond acceptable corporate thresholds.
Shifting Motivations Away From Cost Reduction
Traditional modernization strategies relied on the promise of escaping legacy mainframes to lower operational overhead. That calculus has shifted as infrastructure environments grew more heterogeneous. Kyndryl data indicates that cost reduction and legacy escape no longer serve as the primary drivers for system updates. Instead, tightening sovereignty requirements and the resource demands imposed by modern artificial intelligence workloads force enterprises to demand better operational cohesion across legacy and modern components alike.
This complexity exposes a technical blind spot across the enterprise landscape. Fewer than 10% of organizations express full confidence in their mapping of system dependencies between legacy codebases and newly deployed infrastructure. Without precise dependency maps, attempting broad refactoring initiatives introduces severe operational risks that can take down core transactional databases.
The Role of Agentic AI in Mapping Complex Environments
Autonomous AI agents enter the modernization cycle not to replace underlying mainframes directly, but to expand tangible capabilities across messy technical estates. Rather than rewriting decades-old codebases in a single migration event, early deployments of agentic models focus on administrative and preparatory tasks. These tasks include mapping hidden code dependencies, generating documentation, and producing initial code translations.
These processes have historically resisted automation due to their error-prone nature and the sheer volume of manual verification required by senior software engineers. While agentic AI shows early promise in accelerating these tasks, the low 10% adoption rate demonstrates that enterprises are treating autonomous tools with caution. The historical failure rate of past IT overhauls creates a heavy burden of proof for any new software category claiming to solve legacy debt.
Organizations evaluating these tools must balance the cost of autonomous model execution against the labor savings of automated documentation and dependency mapping. Model inference costs scale directly with codebase size, making brute-force repository analysis economically unviable for smaller IT budgets. Until efficiency benchmarks improve across large-scale code repositories, enterprise adoption will likely remain constrained to incremental tasks rather than wholesale architectural replacement.
Related coverage
- Microsoft’s File History Fix Reveals the Flaw in Treating Legacy Backup Tools as Obsolete
- Okta-Led Alliance Proposes AI Agent Kill Switch to Secure Businesses
- Password Managers vs Authenticator Apps: Security Trade-offs Clarified