P2P

82 The early experimentation centered on model benchmarks. Today, the focus has moved toward architecture, context, and governance. FROM MODEL PERFORMANCE TO ARCHITECTURAL CONTROL Enterprise AI discussions have shifted rapidly since last spring's Peer to Peer publication. The early exper- imentation centered on model benchmarks. Today, the focus has moved toward architecture, context, and governance. The differentiator is no longer which model responds most effectively in isolation, but how a chosen model connects to enterprise systems, retrieves authoritative context, and operates within governance boundaries. The implication is structural, commercial, and strate- gic, with value migrating from the model layer to the orchestration layer. MCP sits at that orchestration boundary and is not intended to just replace APIs. Instead, according to the MCP project's own definitions, it standardizes how models express intent and invoke capabilities across systems. In modular architectures, stable interfaces often determine where advantage consolidates, as described 25 years ago by Baldwin and Clark in "The Power of Modularity." If MCP or a similar standard becomes widely adopted, it may function as such an interface, which may explain why it is getting so much attention. As with all hype, it can run a bit wild, and most sea- soned technical leaders know they need to temper their excitement. Because the protocol alone is not what will leapfrog them ahead or build the moat that shores up the firm's future. With MCP, the real moat lies behind it. THE EMERGING COMPETITIVE MOAT: THE GOVERNED KNOWLEDGE LAYER Law firms have historically been competing for exper- tise embodied in people and documents. As a conse- quence, firms have constantly been, and still are today, raiding each other for talent to build out their capabili- ties, IP, and capacity. But when moving to an AI-centric environment, exper- tise must increasingly be structured, governed, reusa- ble, and retrievable, and cannot rely solely on experts who may be poached on a whim. Instead, the future competitive moat for law firms is likely to be their governed knowledge layer, which governs and provides know-how across: • Structured precedents • Codified playbooks • Tagged matter experience • Clear provenance and authorship • Metadata discipline • Identity-linked access control Established research on modular systems, such as "The Power of Modularity," has long shown that value ac- crues for those who control stable interfaces and rules governing system interaction. And in larger connected ecosystems, competitive control tends to concentrate at the layer that structures and orchestrates coordination across participants, which Jacobides more recently dis- cussed in "Towards a Theory of Ecosystems" in 2018.

• Cover