In this keynote, we'll explore the potential of Large Language Models (LLMs) in unlocking legacy systems and rethinking the connective tissue between legacy datasets and systems with new experiences. Traditionally, enterprises have relied on costly and time-consuming solutions to abstract legacy technology or modernize legacy systems. However, LLMs offer us the change to reimagine this enterprise architecture. By leveraging an LLMs ability to understand APIs, systems, and call tools, LLMs can generate the required structured output to interact with dozens of legacy systems. In this talk, we will discuss the ways LLMs can augment and replace enterprise gateways, reducing the need for custom software development and middleware solutions. We'll also examine the critical role of inference speed in enabling the deployment of LLMs in new and unique ways and explore the potential for LLMs to become a core tool in the enterprise.