Salesforce Agent Fabric adds evals, a kill switch and hard AI budgets
Sorted AI28 OctSalesforce says Agent Fabric, its control plane for agents across platforms, now gets Agent Evals and API Evals (beta in November 2026, generally available January 2027), plus advanced cost controls that are generally available today. Agent Evals can generate tests, check live quality against thresholds, and trigger interventions including an Agent Kill Switch (generally available today; triggered kill switch in January 2027). FinOps Dashboard, Model Wallet and Cost Optimization Policies are generally available; Salesforce says the policies can cut token usage by up to 75% in its test scenarios via caching, compression and response trimming. Salesforce also says Agent Fabric handles over 1.2 billion requests daily.
Why it matters: If you run customer or ops agents on more than one platform, put a quality threshold and a hard spend cap in place before you add another agent, so a silent failure or token spike can be stopped without waiting on a ticket.
Salesforce ↗AI-assisted, reviewed by Sorted AI
▲ 0
0 comments