AI’s most important protocol is getting a little bit easier to use

The economic value here is not in the protocol itself; it is in who gets to sit at the control points once agentic workflows become interoperable. That favors vendors that own identity, policy, audit, and observability layers more than pure model providers: MSFT, SNOW, NOW, OKTA, PANW, and CRWD all have a path to more usage if enterprises standardize how agents touch internal systems. The second-order effect is that interoperability typically lowers switching costs for customers, which increases total AI adoption but reduces the moat of any single closed ecosystem.

The near-term read-through is modest because standards rarely move revenue immediately; the first tangible monetization usually shows up in security, logging, and governance spend before it appears in headline AI budgets. The likely losers are custom integration shops and services-heavy implementers such as ACN, CTSH, and EPAM, where every standardized connector removes billable labor. Over 6-18 months, the bigger risk is that MCP commoditizes the middleware layer and shifts bargaining power from app vendors to platform vendors with the largest distribution.

The contrarian point is that consensus may be over-focusing on AI model competition and underestimating protocol adoption as a re-rating event for enterprise software plumbing. If MCP becomes the de facto rail, the market may pay up for the companies that can enforce trust and governance, not for the companies that merely claim to be AI-enabled. The thesis breaks if enterprise buyers keep these pilots isolated, or if hyperscalers and app vendors fragment the standard with competing interfaces and private connectors.