What's Happening?
Salesforce has announced significant expansions to its Agent Fabric platform, introducing Agent Evals, API Evals, and advanced cost controls. These new features are designed to provide CIOs and AI leaders with proactive control over enterprise AI agents,
preventing issues like silent failures, inaccurate responses, and runaway costs. Agent Fabric, which processes over 1.2 billion requests daily, aims to move organizations from 'cross-your-fingers AI' to continuous control. Agent Evals will automatically generate tests and evaluate agent performance against quality thresholds, triggering interventions like the Agent Kill Switch when results fall short. API Evals will test APIs for breaking changes before they reach production. Advanced cost controls, including the FinOps Dashboard and Model Wallet, will offer budget owners a unified view of AI spending and hard limits to prevent cost overruns, with Cost Optimization Policies potentially cutting token usage by up to 75%.
Why It's Important?
The expansion of Agent Fabric is crucial for enterprises scaling AI, as it addresses critical challenges related to agent performance, reliability, and cost management. With fewer than half of organizations currently evaluating agents on live traffic, the risk of AI agents producing harmful results or incurring excessive costs is substantial. Salesforce's new capabilities provide a much-needed framework for continuous evaluation and automated intervention, ensuring that AI deployments are both effective and financially responsible. This move will instill greater confidence in businesses to adopt and scale AI agents, knowing that robust governance and cost controls are in place. It also highlights the growing complexity of managing AI at an enterprise level and the necessity for specialized tools to oversee these advanced systems.
What's Next?
Agent Evals and API Evals are scheduled for beta release in November 2026, with general availability expected in January 2027. Custom Agent Scanner is currently in beta and will be generally available in January 2027. The Agent Kill Switch, Open Policies for Federated API Gateways, FinOps Dashboard, Model Wallet, and Cost Optimization Policies are generally available today. These phased rollouts indicate Salesforce's commitment to progressively enhancing AI governance. Customers will be able to leverage these tools to automatically detect and resolve issues, manage AI spending, and ensure agents operate within defined quality and budget parameters. The continuous development and release of these features will be critical for enterprises looking to responsibly scale their AI initiatives.
Beyond the Headlines
Salesforce's Agent Fabric expansion signifies a maturing phase in enterprise AI adoption, where the focus is shifting from initial deployment to ongoing management, reliability, and cost optimization. This development underscores the inherent challenges of integrating AI into critical business operations, particularly the need for robust oversight to prevent unintended consequences. The introduction of features like the 'Agent Kill Switch' highlights the recognition that AI systems, while powerful, require human-defined guardrails and intervention mechanisms. This trend will likely lead to increased demand for AI governance solutions across the industry, prompting other vendors to develop similar capabilities. Ultimately, it reflects a growing understanding that successful AI implementation is not just about technological prowess but also about responsible and controlled deployment.
