Transforming legacy from within: Agentic AI and the evolution of enterprise systems
By Naresh Duddu, AVP and Global Head – Modernization, Infosys
Enterprise IT budgets bleed through one crack, that of keeping legacy systems chugging along. No surprises there, as most large enterprises have core legacy systems that carry within them irreplaceable institutional knowledge. These systems are often the backbone on which the enterprise runs. Abandoning them could affect the business in the long term. But the very things that make these old systems so essential also make them a huge risk. As per industry research, over 75% of technology professionals are concerned about security vulnerabilities in legacy systems. Tech leaders are no longer arguing over whether they should upgrade. Instead, they are trying to do it without breaking everything.
Enterprises have been trying to modernize legacy systems for years using different methods, but with limited success. Rewriting systems from scratch is often not viable due to several reasons such as prohibitive costs and timelines. Re-architecting exposes hidden dependencies that affect progress. Incremental, module-by-module changes slow down the momentum to reach the required scale. While generative AI tools help accelerate code translation and modernization efforts, they typically produce output that replicates legacy behavior without resolving the underlying architectural debt. AI agents change that. They can orchestrate significant portions of the modernization lifecycle, from discovery and analysis to code transformation, testing, and remediation, while keeping architects and business stakeholders in key decision-making roles. All of this can reduce manual effort significantly and accelerate execution.
Understand the code’s intent, not just its structure
Generative AI has accelerated code translation, but translation alone is not modernization. Most tools simply convert legacy code into a newer language while preserving the same monolithic architecture, tightly coupled components, and hardcoded dependencies. The result is often a modernized codebase that delivers little improvement in terms of agility, scalability, or maintainability. Basically, organizations end up with a shinier version of the same problem.
Agentic AI changes this by automating the broader modernization lifecycle. Specialized agents can analyze legacy applications, recommend service boundaries, generate candidate cloud-native implementations, and accelerate validation through automated testing. By automating these labor-intensive activities, organizations can reduce modernization effort, accelerate delivery, and enable architects to focus where it counts: making architectural calls, deciding what to modernize first, and managing risk. Tasks previously too complex or expensive to automate at scale such as database migrations and re-engineering back-end to front-end layers are now executable with greater speed and reduced dependence on subject matter experts.
Agentic AI also opens new possibilities for automating the “white spaces”. These are processes that were never fully digitized because they relied on human judgment. Agentic AI can help automate/assist judgment-intensive processes by combining contextual reasoning with human oversight.
Emergence of the ‘New Enterprise Model’
By embedding AI Agents into existing legacy systems, enterprises are now evolving from rigid, process-bound models to dynamic, AI-augmented operations that scale and evolve with the needs of the business. This evolution occurs through two primary vectors:
Firstly, AI-accelerated workflows can augment core business operations through multi-agent coordination, real-time insights, and decision support, helping teams improve speed, accuracy, and efficiency within the legacy ecosystems.
Secondly, one with Agentic AI infusion within the existing enterprise applications powered by AI-native data modernization replaces keyword based retrieval with NLP-driven semantic intelligence, delivering real-time insights, personalized automation, and scalable intelligence across enterprise.
Governing the transition: A necessary discipline
The same autonomy that makes agentic AI effective in modernization also introduces governance requirements that enterprises must address before scaling. A credible Agentic AI modernization strategy must be paired with appropriate governance: defined boundaries on what agents are permitted to do, human-in-the-loop checkpoints for high-stakes decisions, audit trails for agent actions, and continuous monitoring for unexpected behavior. Begin with targeted, high-impact use cases where agent decisions are transparent, measurable, and reversible. Demonstrate results, then scale with confidence.
The question for decision-makers is no longer whether to use AI in modernization, but how to do so with discipline. Success depends on prioritizing the right use cases, establishing robust governance, and building the organizational capabilities needed to sustain change.
While traditional modernization delivers incremental improvements, agentic AI shifts the paradigm from code assistance to autonomous, outcome-driven execution. With the right guardrails, enterprises can accelerate modernization, reduce technical debt, and transform legacy systems into a foundation for an adaptive, AI-native future.