Salesforce has launched the Trusted Enterprise AI Harness, a comprehensive architecture designed to enable AI agents to operate with trust, context, and control across enterprise systems. The announcement, made on September 11, 2026, introduces six foundational capabilities—Trusted Context, Trusted Agency, Trusted Action, Trusted Governance, Trusted Security, and Trusted Models—all delivered through a composable architecture that builds on existing Salesforce technologies like Data 360, Informatica, MuleSoft, Agent Fabric, Tableau, Agentforce, Salesforce Guardian, and the Salesforce Platform (salesforce.com).
At the heart of the offering is a new AI Control Plane, which provides enterprises with a centralized interface to discover, register, manage, and monitor AI agents and models across both Salesforce and third-party systems. It supports lifecycle management, policy enforcement, performance evaluation, cost control, and observability—addressing the growing complexity of multi-agent deployments (salesforce.com).
The Trusted Enterprise AI Harness is built to be composable and open, allowing organizations to adopt only the components they need and integrate with external models and systems. Salesforce emphasizes that enterprises don’t have to rely solely on its stack—capabilities are accessible via MCP, APIs, Skills, and plug-ins, and can extend into platforms like Claude, Slack, Microsoft Teams, and Agentforce as part of its broader AIforce strategy (salesforce.com).
This launch comes ahead of Salesforce’s Dreamforce conference (September 15–17, 2026), with general availability expected to begin in early fiscal year 2028. Many of the underlying technologies are already available to existing customers, with upgrades and new capabilities rolling out over time (salesforce.com).
Why it matters: Enterprises today often run multiple agent orchestration platforms—VentureBeat Intelligence reports an average of 3.1 per organization—and need a unified control layer to manage them securely and efficiently. Salesforce’s AI Harness addresses this fragmentation by offering a trusted, enterprise-grade foundation for AI agents to understand context, act reliably, and remain governed across diverse environments (venturebeat.com).