The promise of industrial-era stability in the digital age is officially dead. Narayana Murthy, the billionaire co-founder of Infosys, recently asserted that “legacy systems” would act as a bulwark against the AI-driven purge of tech jobs. Murthy’s logic suggests that the complexity and age of existing corporate infrastructure require a human touch that artificial intelligence cannot replicate.
He is fundamentally wrong. The assumption that the “maintenance” of antiquated codebases is a protected human sanctuary ignores the rapid evolution of Large Language Models (LLMs) specifically tuned for legacy translation and refactoring.
For the African tech ecosystem—which often acts as the outsourced engine for these global legacy systems—Murthy’s optimism isn’t just misplaced; it is dangerous. It encourages a complacency that could leave an entire generation of developers obsolete within the next 36 months.
—
## The Myth of the “Un-automatable” Legacy System
Murthy’s argument hinges on the idea that legacy systems are “spaghetti code” so tangled that only a human brain can navigate the nuance. In the traditional outsourcing model, firms like Infosys and TCS deployed thousands of junior engineers to manually patch, debug, and maintain these systems.
This labor-intensive model is precisely what AI is designed to destroy. Modern AI tools are no longer just writing fresh code; they are being trained on COBOL, Fortran, and early-stage Java—the very languages Murthy claims provide a safety net.
> “Maintenance is the low-hanging fruit of the AI revolution,” says Dr. Arishe Ogundipe, a lead systems architect based in Lagos. “AI doesn’t get bored scanning ten million lines of 20-year-old code. It maps dependencies in seconds—work that would take a human team six months.”
The reality is that “legacy maintenance” is a data-processing task. And in the world of data processing, the human is always the weakest link.
—
## The Data: The Numbers Behind the Replacement
The shift is already reflected in the hiring patterns of global tech giants. While Murthy speaks of stability, the financial data suggests a structural retreat from human-heavy maintenance models.
– **Headcount Stagnation:** Major Indian and African tech hubs have seen a 25% drop in campus recruitment for entry-level maintenance roles over the last fiscal year.
– **Refactoring Efficiency:** Internal benchmarks from firms utilizing GitHub Copilot and Amazon CodeWhisperer show a 40% increase in speed for legacy migration tasks.
– **Cost Arbitrage:** An AI-driven maintenance suite costs approximately 5% of the annual salary of a junior offshore developer, while operating 24/7 without benefits or downtime.
The “bench” model—where outsourcing firms keep thousands of developers ready for legacy projects—is becoming a liability. Investors are no longer rewarding firms for the size of their headcount, but for the efficiency of their automation.
—
## The Automation of the “Boring” Work
Murthy’s defense rests on the belief that legacy systems are too “fragile” for AI. This ignores the emergence of “automated refactoring” startups that specialize in moving legacy bank systems to the cloud.
These tools do not just patch the old code; they rewrite it into modern, efficient languages. The human role in this process is shrinking from “creator” to “reviewer.” One senior engineer can now oversee the work previously assigned to a team of twenty.
### Why the “Human-in-the-loop” is temporary:
– **Accuracy Rates:** AI models are reaching 90% accuracy in translating obsolete code into modern syntax.
– **Syntactic Analysis:** AI can identify security vulnerabilities in legacy code that human eyes have missed for decades.
– **Scalability:** An AI can “learn” a proprietary, 40-year-old internal system in hours. A human trainee takes months to reach billable proficiency.
—
## Impact: The Looming Crisis for the African Developer
For the rising tech hubs in Nairobi, Lagos, and Kigali, Murthy’s narrative provides a false sense of security. Africa has positioned itself as the new frontier for Western tech outsourcing, often handling the “maintenance” and “support” roles that Murthy mentions.
If African developers believe that legacy maintenance is a career-long insurance policy, they will stop upskilling. They will continue to learn “maintenance” rather than “architecture” and “AI integration.”
> “We are seeing a disconnect between what industry veterans are saying and what the capital is doing,” says a venture analyst specializing in African FinTech. “Backing legacy maintenance as a job creator is like backing the horse-and-buggy industry because the new cars still need some roads paved. The engine has changed.”
The transition will be brutal for those who do not adapt. The middle-class dream built on outsourced debugging is evaporating.
—
## The Displacement is the Strategy
The goal of modern enterprise tech is not to maintain legacy systems forever, but to eliminate them. AI is the catalyst that allows corporations to finally kill off their “legacy” baggage.
When Murthy says legacy systems will save jobs, he is defending a business model—the billable hour—that is fundamentally at odds with the efficiency of AI. For the worker, the “maintenance” role is not a fortress; it is a sunset industry.
The next five years will not be defined by humans maintaining the old, but by AI replacing the old with the new. To suggest otherwise is to ignore the fundamental economics of the current technological shift. The safety net that Murthy describes is made of thin air.
