Cisco Bets on AI to Rewrite Nearly All of Its Legacy Code Within Two Years
Cisco is preparing for one of the most ambitious software modernization efforts in its history: using artificial intelligence to help rewrite 98% of its legacy code within the next two years.
The bold plan comes from Paddy Krishnamurthy, a 27-year-old former startup founder who joined Cisco after the company acquired his real-time translation startup, EzDubs, in 2025. According to Cisco President and Chief Product Officer Jeetu Patel, Krishnamurthy had been placed several levels down in the organization and was working on projects that were not central to Cisco’s long-term strategy.
That changed after a conversation at Cisco’s San Francisco office. Krishnamurthy proposed an aggressive AI-driven approach to modernizing the company’s massive codebase, initially suggesting that the work could begin with just himself and one other person. Instead of dismissing the idea as too ambitious, Patel gave him unlimited access to AI tokens and encouraged him to push the experiment as far as possible.
Krishnamurthy has reportedly set a four-week window to prove that the approach can work. If successful, the broader goal is to overhaul almost all of Cisco’s legacy code in 24 months.
For a company that has spent more than 40 years building networking hardware, enterprise software, and infrastructure tools, the scale of the challenge is enormous. Legacy code is often deeply connected to older systems, internal tools, customer products, and long-standing engineering workflows. Rewriting it is usually slow, expensive, and risky.
Cisco’s experiment suggests a different path: using AI coding tools to accelerate software refactoring, reduce technical debt, and help engineers modernize systems faster than traditional development methods allow.
Patel has described the small team as a catalyst rather than a replacement for Cisco’s wider engineering workforce. The idea is not for two or three people to independently rebuild the company’s software foundation. Instead, the early team will test what is possible, while Cisco’s engineering teams will ultimately own, review, and refine the final results.
This approach could reshape how large technology companies think about legacy code modernization. Many established businesses are sitting on decades of software that still works but is difficult to maintain, secure, or adapt to modern demands. AI-assisted development may offer a way to clean up old systems, improve performance, and move faster without starting from scratch.
One possible benefit is the transition toward memory-safe programming languages. These languages are designed to reduce common memory-related bugs, which can lead to crashes, vulnerabilities, and security risks. For a company like Cisco, whose products are used across enterprise networks worldwide, improving software safety and reliability is a major priority.
Patel has acknowledged that not every AI experiment will succeed. Some projects may fail, and some ideas may not scale. But he believes the potential payoff is too significant to ignore. If AI can help engineers rewrite, test, and improve code at a much faster pace, it could change the speed at which enterprise software is built and maintained.
The initiative also reflects a broader shift in how Cisco views young talent. Patel has pointed to younger employees and interns as sources of fresh thinking, especially around artificial intelligence. In his view, people with less industry experience may be more willing to question old assumptions and experiment with new workflows.
He has described interns as “reverse mentors,” meaning they are not simply brought in for routine work, but to expose experienced teams to new ways of thinking. One example involved a Stanford student who earned an internship after consistently updating Patel on what she was learning about AI. Once inside Cisco, she reportedly worked on a redesign that more experienced staff had struggled to solve.
This mindset is becoming more common across the tech industry. Some leaders are increasingly interested in hiring people with fewer years of experience because they may be less attached to traditional processes. Others are looking for candidates who can adapt quickly, experiment with AI tools, and challenge how work has always been done.
Cisco’s plan to use AI to rewrite 98% of its legacy code is still in its early stages, and the final outcome remains uncertain. But the move shows how seriously major technology companies are taking generative AI and AI-powered software development.
If the effort succeeds, it could become a major example of how artificial intelligence can help large enterprises modernize old codebases, improve software security, and speed up product development. It may also prove that the future of engineering is not just about bigger teams, but about smaller groups using powerful AI tools to unlock change across an entire organization.
