Mahesh Kumar Goyal, senior data and AI expert at Google LLC, urged tech leaders to rethink how legacy systems that are often dismissed as outdated are actually the most valuable foundation for building successful AI solutions. Addressing industry experts during a recent virtual interaction, Goyal cautioned against the sector’s growing obsession with “greenfield innovation” while ignoring decades of accumulated business knowledge.Referring to his recent participation at Google Cloud Next 2026, Goyal said the conversation around “Agentic AI” is accelerating globally, but many organizations are moving in the wrong direction. “We’re often tempted to build the future from scratch,” he told the audience, “but the real future is buried in the systems companies already have.” If you ignore them, your AI will lose context and fail in the real world.“Goyal explained that most enterprises separate their AI initiatives from their modernization efforts, a mistake he described as an “architectural lie.” “Your modernization strategy and your AI strategy are the same strategy,” he said. “If you treat them differently, you’ll create an impressive demo, not an impressive system.”Sharing insights from his experience, Goyal recounted a large-scale cloud migration project where an advanced system failed during an AI pilot. The reason, he said, was simple: the critical business logic embedded in the old systems had been lost. “Legacy systems are not technical debt,” he emphasized. “They are encoded institutional memory. You can rewrite code, but you can’t rewrite knowledge.”He advised organizations to develop frameworks that bridge old and new technologies rather than treating them as mutually exclusive. According to him, emerging approaches such as Model Context Protocol (MCP) and GraphRAG are enabling engineers to connect AI models to historical systems, unlocking insights that would otherwise remain hidden. “Consider the MCP as a peace treaty,” he said. “This allows advanced AI agents to communicate without destroying legacy systems.“At the same time, Goyal warned about the dangers of poorly integrated AI systems. While basic chatbots may cause minor problems, he noted that AI agents operating on incomplete or poorly understood legacy data can lead to serious financial consequences. He cited an example where graph-based analysis revealed unknown systems handling large vendor transactions, systems that senior leadership was unaware of.For industry leaders, Goyal’s message was clear: technical expertise alone is not enough. “The real challenge is not the technology, it’s the architectural courage,” he said. “You have to be willing to question how systems are designed and where knowledge actually resides.”Concluding his address, he offered a practical test for executives running these organizations: “Take the top three AI projects and the top three modernization projects in your company. Put them on the same slide. If you can’t connect them, you have no strategy.”As companies around the world race to adopt AI, Goyal emphasized that those who succeed are not necessarily those with the most advanced models, but rather those who learn to extract intelligence from existing systems.
