Google and Microsoft Back ARD Protocol to Counter AI Superapps

Google, Microsoft, Salesforce, Snowflake, and ServiceNow have backed a new open software standard called Agentic Resource Discovery (ARD), designed to make AI models interchangeable components inside existing enterprise applications.

Key Points:

• ARD extends Anthropic's Model Context Protocol (MCP)—the standard that lets AI systems connect to external databases—by adding a DNS-like catalog and registry system for dynamically discovering AI tools across a network. In plain terms: ARD is like a phone book for AI tools inside a company—any AI agent can look up what tools are available and use them, without being locked into one vendor's ecosystem.

• The standard allows enterprise applications like Salesforce or Microsoft 365 to route AI queries to any compliant model, keeping users inside the host app rather than moving to a standalone AI workspace.

• OpenAI and Anthropic are not signatories to the initial ARD coalition—reflecting the strategic conflict: both companies want their superapps to be the primary workspace, not a background service inside existing apps.

• The ARD/superapp divide is shaping up as the central strategic battle of enterprise AI for 2026-2027.

If ARD becomes the enterprise standard, it gives traditional software vendors the ability to commoditize AI models—preventing any single lab from capturing the workspace. For IT architects, ARD fluency and its underlying MCP foundation are becoming essential for designing vendor-neutral enterprise AI pipelines.

Why It Matters: The ARD vs. superapp battle will determine whether AI labs capture enterprise value or traditional software vendors do. Understanding this divide is essential for navigating AI procurement and architecture decisions.