Unlocking Enterprise Knowledge Hidden in Legacy Systems

For many large enterprises, the greatest barrier to modernization is not outdated technology, but the knowledge embedded within it. Decades of software development, business rules, and operational expertise often remain buried across millions of lines of code, scattered documentation, and the experience of a handful of specialists.

This challenge recently surfaced at one of the world's leading, U.S.-headquartered automotive services and technology companies, where engineers frequently spent 40-60% of their time searching for information before they could safely modify software systems. Understanding system dependencies, evaluating risks, and identifying the impact of even minor software changes often required navigating layers of business logic accumulated over decades. The experience highlights a growing reality for enterprises navigating AI-driven transformation: before AI can help organizations move faster, it must first understand how their systems actually work.

 

According to Mr. Pham Minh Tuan, Executive Vice President and FPT Software CEO, FPT Corporation, legacy systems should be viewed as valuable business assets rather than technological burdens. “These systems represent decades of accumulated operational knowledge and business logic. They are not infrastructure liabilities, but strategic assets that enterprises cannot afford to lose. Our goal is to help organizations transform that complexity into a competitive advantage in the AI era,” he explained.

 

To help enterprises unlock and operationalize the hidden knowledge embedded within complex software environments, FPT developed Flezi Metis, an AI-powered platform that connects source code, technical documentation, and operational data into a unified knowledge graph. By creating a single source of truth across enterprise systems, the platform enables AI to understand how applications, processes, and information are connected, providing the context needed for more accurate analysis and decision-making.

 

In an implementation involving approximately one million lines of code, Flezi Metis reduced knowledge-search time by around 50% while improving engineering task completion time by approximately 30%. Beyond productivity gains, the initiative demonstrates how AI can help organizations transform expertise locked inside complex systems into knowledge that can be shared and reused across the enterprise.

 

 

While many organizations have adopted AI assistants and coding copilots to improve productivity, traditional AI tools often struggle to understand the broader context needed to deliver accurate analysis and recommendations, given increasingly large and fragmented enterprise systems. FPT has observed this challenge across major markets such as North America, Europe, and Japan. IDC estimates that approximately 80% of medium and large enterprises in Japan continue to rely on legacy systems, highlighting the widespread need to modernize complex technology environments while preserving valuable institutional knowledge.

 

Flezi Metis was developed to bridge this gap. The solution connects source code, technical documentation, and operational data into a unified knowledge graph,  AI to understand relationships across systems and business processes, helping engineering teams identify dependencies, assess potential impacts, and pinpoint affected areas before implementing changes. Data remains within customers' existing infrastructure, supporting stringent security and compliance requirements.

 

 

The challenge of managing fragmented knowledge is becoming increasingly important as organizations seek to scale AI adoption. A global study commissioned by FPT and conducted by Forrester Consulting found that while 51% of organizations allocate at least 5% of their IT budgets to AI, only 26% consider themselves advanced in operationalizing it. The research also identified integration complexity (41%) and data silos (38%) as the leading barriers to scaling AI initiatives.

 

As enterprises move beyond isolated AI pilots, they are increasingly seeking partners capable of supporting AI implementation across the full lifecycle. According to the study, nearly half of organizations (48%) prioritize partners that can engineer, deploy, and operate AI systems at scale while ensuring governance, security, and integration with existing technology environments.

 

To meet this demand, FPT's enterprise AI ecosystem FleziPT is designed to help organizations modernize legacy environments, unlock institutional knowledge, and accelerate AI adoption across business functions. The ecosystem is further strengthened by FPT's global AI-augmented workforce, AI Factories in Vietnam and Japan, and collaborations with leading technology partners, enabling enterprises to scale AI transformation with greater speed, governance, and operational impact.

 

Complementing the ecosystem, FPT recently introduced CASAN, a structured AI maturity framework that guides organizations from fragmented experimentation toward AI-native operations with stronger governance and measurable outcomes.

 

 

According to Dr. Nguyen Xuan Phong, Chief AI Transformation Officer at FPT Corporation: “In complex enterprise environments, the greatest risk lies in knowledge locked inside source code. What once existed only in the experience of a few specialists must become an organizational capability.”

 

While the automotive sector provides a clear example of this challenge, similar issues are emerging across industries including banking, insurance, and the public sector, where organizations continue to rely on complex legacy environments while accelerating AI adoption.

 

Technology modernization is no longer just a software challenge. It is increasingly a knowledge challenge. As organizations seek to scale AI, the systems that power their businesses must become easier to understand, not just easier to automate. The companies that succeed will be those that can transform decades of accumulated expertise from information locked inside code into knowledge that can be shared, analyzed, and acted upon across the enterprise. In that sense, the future of AI may depend as much on organizational understanding as on technological advancement.