Salesforce previews Trusted Enterprise AI Harness with new AI Control Plane
Salesforce previewed its Trusted Enterprise AI Harness on September 10, 2026, introducing a new architecture designed to give AI agents shared business context, controlled access to enterprise systems, and centralized oversight. The announcement positions Salesforce as more than an application provider in the agentic AI market, with the company aiming to become the control layer connecting models, agents, data, workflows, and governance.
Salesforce Trusted Enterprise AI Harness brings six AI capabilities together
The architecture combines six capabilities: trusted context, trusted agency, trusted action, trusted governance, trusted security, and trusted models. Salesforce says the framework is built across technologies from Data 360, Informatica, MuleSoft and Agent Fabric, Tableau, Agentforce, Salesforce Guardian, and the Salesforce Platform. The company described the approach in its September 10 announcement of the Trusted Enterprise AI Harness.
The design addresses a practical problem with enterprise AI deployments. Customer information, inventory, contracts, policies, workflows, and previous interactions usually sit across separate systems. An agent may be able to generate a plausible answer, but reliable execution requires access to the right data, business rules, permissions, and actions at the same time.
Salesforce is therefore treating enterprise AI as an infrastructure problem rather than only a model problem. Trusted context is intended to provide a shared view of the customer and business, while trusted action connects agents to applications, APIs, workflows, and business processes. Governance and security are intended to control the data agents use and the actions they can take.
New Salesforce AI Control Plane targets agent oversight
The announcement also introduces an AI Control Plane for managing agents and AI capabilities across an enterprise. Salesforce says the control layer will let organizations discover and register agents, establish identity and policy, manage lifecycles, evaluate performance, observe behavior and outcomes, and control costs.
That focus on centralized oversight is significant because agent deployments rarely remain confined to one CRM workflow. Different departments may use separate agents, models, and automation tools, creating fragmented permissions, inconsistent monitoring, and limited visibility into operational cost. A shared control plane could give IT, security, data, and business teams a common way to manage those systems.
SiliconANGLE reported that Salesforce is presenting the AI Control Plane as a way to improve the reliability and security of customer AI agents, while many of the underlying technologies are already available through Salesforce cloud services. The broader direction is clear: Salesforce wants customers to manage both Salesforce and third-party AI from a common operating layer.
Salesforce is previewing an architecture, not a finished product suite
The immediate limitation is availability. Salesforce says many technologies forming the foundation of the Enterprise AI Harness are available today, but new capabilities and the unified experience are planned to begin rolling out in early fiscal FY28. Pricing, packaging, upgrade paths, and regional availability will be announced closer to general availability.
That makes the September 10 announcement a strategic preview rather than a complete product launch. VentureBeat described the offering as a planned platform that will not be fully available to all customers until early next year. Customers evaluating the announcement should therefore distinguish between capabilities they can use now through existing Salesforce products and the unified harness experience that remains on the roadmap.
Salesforce shifts Agentforce toward enterprise AI infrastructure
The announcement expands Salesforce’s Agentforce strategy from deploying individual agents to managing the context and controls those agents depend on. Existing customers could eventually gain an upgrade path that makes their Salesforce data, metadata, business logic, permissions, and workflows reusable across more AI experiences, including third-party models and agent surfaces.
That positioning could strengthen Salesforce’s role as companies connect AI to operational systems. It also raises the execution challenge. The value of the harness will depend on how consistently Salesforce integrates its product portfolio, how openly the capabilities work with external models and agents, and whether customers can adopt the controls without creating another layer of complexity.
For now, the most important development is the architectural shift. Salesforce is treating trusted context, governed action, and centralized AI oversight as a unified enterprise platform, with the full experience expected to develop through fiscal FY28.
