Salesforce has unveiled two new tools, Enterprise AI Harness and AI Control Plane, designed to assist developers in building and managing artificial intelligence agents. The Enterprise AI Harness focuses on enhancing the reliability and security of these agents, with a rollout scheduled to begin in February 2027, marking the start of the company's 2028 fiscal year.
The Enterprise AI Harness consists of six feature collections, including Trusted Context for managing customer records and product data. It also introduces data quality controls, such as a lineage tool, and cybersecurity features that allow teams to regulate dataset access. Additionally, the platform includes intelligent model routing to automatically direct requests to the most cost-effective and accurate large language models.
To address the non-deterministic nature of large language models, the new offering provides reliability controls and simplifies connections to external applications. Meanwhile, the AI Control Plane is designed for centralized monitoring, allowing administrators to track agent performance, manage inference costs, and ensure compliance with cybersecurity policies.
Rohan Kumar, president and Chief Platform and Engineering Officer at Salesforce, stated that the company aims to help enterprises securely integrate proprietary customer data into AI workflows. Many of the underlying technologies for these offerings are already integrated into existing Salesforce cloud services.
