Salesforce unveils Enterprise AI Harness and AI Control Plane

Salesforce has previewed two new offerings designed to help enterprises build more reliable and secure AI agents while maintaining centralized oversight of their performance and costs.

The company announced the Enterprise AI Harness and AI Control Plane, both aimed at addressing critical challenges organizations face when deploying AI agents across their operations. According to Salesforce, many components of these offerings already exist within its cloud services, with additional features rolling out starting in February 2026, the beginning of the company's fiscal 2028.

Six feature collections for agent reliability

The Enterprise AI Harness comprises six main feature collections, half of which focus on managing the business data that AI agents rely on to function effectively.

The first collection, called Trusted Context, aggregates data assets including customer records and product availability information. It will also support real-time signals—buyer actions that require immediate adjustments to outreach strategies.

A second set of features ensures data accuracy through multiple quality controls, including data lineage tracking. Security teams will gain granular control over which agents can access specific datasets and how they can use that information.

The third data-focused capability addresses a common enterprise practice: using multiple large language models to balance performance and cost. Organizations typically route complex tasks to frontier models while directing simpler queries to smaller, cheaper alternatives. Salesforce's intelligent model routing mechanism will automate this process based on accuracy and cost parameters.

Addressing nondeterminism and integration

Because LLMs are nondeterministic—meaning identical prompts can produce different outputs—the Enterprise AI Harness will pair agents with deterministic reliability controls. The offering will also simplify connecting agents to external applications, a frequent integration challenge.

Centralized monitoring and governance

The AI Control Plane provides administrators with a unified view of AI agent performance and inference costs across their organization. It will also enforce cybersecurity policy compliance, addressing governance concerns that often slow enterprise AI adoption.

Why it matters

As enterprises move from experimenting with AI to deploying agents in production environments, they face mounting challenges around data quality, security, cost management, and regulatory compliance. Salesforce's approach recognizes that the competitive advantage in enterprise AI won't come from model access—which is increasingly commoditized—but from how organizations integrate proprietary business context with AI capabilities while maintaining control and trust. These offerings directly address the operational gaps that prevent many companies from scaling AI agent deployments beyond pilot projects.

"Models will continue to change, and intelligence will increasingly be available everywhere," said Rohan Kumar, Salesforce's president and chief platform and engineering officer. "What will differentiate an enterprise is the trusted, proprietary context it brings to that intelligence—starting with the customer—and its ability to securely turn that context into action."

These details were first reported by SiliconANGLE.