Software Architechture AI Day 2026

The Model Context Protocol (MCP) was published a year ago and has gained rapid traction, and for good reasons. For the first time, it provides a standardized way for LLMs to interact with the environment. However, MCP’s goal was never for that environment to be a vast landscape of organizational capabilities; it started with modest goals such as providing access to your local filesystem.

But what does it take to enable AI enablement at scale in large organizations? This presentation focuses on the API side of things and looks at what comes after the basic LLM access to APIs that we can get with MCP. We explore how tool descriptions must evolve, how agents must discover those tools at runtime, and how workflows should be exposed as single, coherent tools. Looking at this we see a progression that many organizations will go through over time, and the sooner you start thinking about this path, the better you can steer it and invest strategically in those areas that are most important for you.