Foolish to think large enterprises will dump legacy IT systems overnight to use AI: Former Infosys CFO Mohandas Pai
Large companies cannot simply discard their decades‑old IT infrastructure and replace it with AI‑driven platforms in a few months, says a former chief financial officer of Infosys. The comment comes as CEOs worldwide promise rapid digital transformation, but the reality on the ground is far more gradual.
Legacy systems and modern AI
Most multinational corporations still rely on mainframes, custom‑built applications, and on‑premise data centers that were installed in the 1990s or early 2000s. These systems manage core functions such as payroll, supply‑chain logistics, and customer records. While they are stable, they were not designed for the massive data flows and real‑time analytics that generative AI models require.
Why the hype feels urgent
Industry analysts and technology vendors have been promoting AI as a shortcut to efficiency, cost savings, and new revenue streams. Press releases often showcase pilot projects that cut processing time by half or generate marketing copy in seconds. Such success stories create pressure on boardrooms to act quickly, sometimes without a clear roadmap for integrating AI with existing platforms.
Insights from a former CFO
The ex‑CFO, who oversaw Infosys’s finance operations during a period of rapid growth, cautioned that “expecting a wholesale dump of legacy code overnight is unrealistic.” He explained that finance teams must first understand the cost structure of current applications, assess regulatory compliance, and ensure data integrity before any AI layer can be added. Rushing the process could expose firms to audit failures, security breaches, or costly downtime.
Legacy applications often use outdated programming languages, proprietary data formats, and rigid interfaces. Connecting these to modern AI services typically requires building middleware, data pipelines, and API gateways. Each of these components adds complexity and demands specialized skills that many IT departments lack. Moreover, AI models need clean, labeled data—a luxury that older systems rarely provide without extensive preprocessing.
Cost and risk considerations
Replacing or refactoring legacy code can cost billions for a global enterprise. The financial outlay includes licensing new software, hiring data scientists, and training staff. In addition, any misstep can disrupt critical business processes, leading to revenue loss and reputational damage. The CFO highlighted that many firms underestimate these hidden expenses, focusing instead on the headline‑grabbing benefits of AI.
The challenge is not limited to any single market. Companies in North America, Europe, and Asia all face similar legacy burdens. In emerging economies, the gap is even wider because many organizations still operate on on‑premise hardware due to limited cloud connectivity. As AI adoption spreads, the disparity between early adopters and those stuck with old systems could widen, affecting global supply chains and competitive dynamics.
Practical strategies for transition
Experts recommend a phased approach. First, organizations should conduct an inventory of existing applications, categorizing them by criticality and modernization potential. Next, they can adopt a “lift‑and‑shift” to cloud infrastructure, which provides a more flexible foundation for AI tools. After that, selective re‑engineering—rewriting only the most AI‑relevant modules—allows firms to test models without overhauling the entire stack. Finally, establishing a governance framework ensures that AI deployments meet compliance and ethical standards.
The CFO’s warning serves as a reminder that technology change is a marathon, not a sprint. While AI will undoubtedly reshape business operations, the pace of change will be dictated by the readiness of underlying IT assets. Companies that invest in incremental upgrades, robust data management, and skilled talent are more likely to reap sustainable AI benefits. Those that chase quick wins without addressing legacy constraints risk costly setbacks.
In summary, the path from legacy IT to AI‑enhanced operations is complex and costly, especially for large enterprises with entrenched systems. A realistic, step‑by‑step plan—grounded in financial prudence and technical feasibility—offers the best chance for success. The industry’s next wave of growth will depend not on overnight miracles, but on steady, well‑managed transformation.