Salesforce adds spending caps and agent evals to Agent Fabric
Announced · Published · By Contractor AI Shop
Salesforce added spending controls to Agent Fabric: a FinOps dashboard and Model Wallet that hard-caps agent spend and stops over-budget requests. Agent Evals and API Evals (beta Nov, GA Jan 2027) can trigger a human-approved shutdown when quality slips; budget controls are available now.
Key facts
- Salesforce says FinOps Dashboard, Model Wallet and Cost Optimization Policies are generally available now; Cost Optimization Policies claim up to 75% token reduction based on Salesforce test scenarios. Salesforce News
- Agent Evals and API Evals enter beta in November 2026 and are planned GA in January 2027; they can trigger interventions including the Agent Kill Switch with human approval where policy requires it. Salesforce News
- VentureBeat (Carl Franzen) reported the Nov beta / Jan 2027 dates, noted the basic Kill Switch was already GA in September, and interviewed Andrew Comstock on cross-platform agent governance. VentureBeat (Carl Franzen)
- WebProNews (Rich Ord) reported the FinOps Dashboard, Model Wallet and Kill Switch as available now, and the evals beta/GA schedule. WebProNews (Rich Ord)
Our take: what it means for contractors
Enterprise-only today, but the idea carries down to a contractor shop: put a hard dollar cap on any AI agent's model spend, and re-test an agent after you swap it to a cheaper model.
What we don't know yet
- Pricing for Agent Evals / Agent Fabric
- Any customer-reported savings (75% is vendor test data)
Sources
- Salesforce News · Salesforce Expands Agent Fabric to Automatically Stop Runaway Costs and AI Agent Failures Before They Impact Customers · 2026-10-27 · Primary
- VentureBeat (Carl Franzen) · Salesforce expands Agent Fabric with AI evaluations to trigger shutdowns when agents fail · 2026-10-27 · Secondary
- WebProNews (Rich Ord) · Salesforce Adds Hard Budget Caps and a Kill Switch to Agent Fabric · 2026-10-28 · Secondary