# Faultprint Operating Diagnostic: Salesforce Agentforce

Run ID: FP-AI-OPERATING-0090
Mode: Operating Condition Mode
Institution: Salesforce
Primary source: Salesforce Agentforce public materials

## Executive Read

Salesforce Agentforce is an enterprise agentic-AI corpus centered on autonomous agents, CRM actions, business data, guardrails, trusted AI, action definitions, and customer workflows. The structural condition is CRM action authority: agents may update records, answer customers, follow up with leads, or execute workflow steps in systems that sales, service, and operations teams treat as authoritative.

## Findings

1. Salesforce describes Agentforce as trusted autonomous AI agents for business workflows.
2. Salesforce public materials describe agents tailored with actions, data, and guardrails.
3. Salesforce Trailhead materials describe guardrails as operational boundaries defining what agents can and cannot do.
4. The exposure is customer-record and workflow action dependency: CRM actions require proof of permissions, grounding, escalation, audit trails, and rollback.
5. Comparable failure anatomy includes customer-facing AI reliance and agentic action cases where generated responses or actions exceeded authorization.

## Falsifiability

The exposure would be downgraded if the assessed corpus included agent action scopes, CRM field-level permissions, grounding logs, guardrail tests, human escalation records, audit trails, and rollback records for erroneous agent actions.

## Boundary

This run does not assert Agentforce causes erroneous CRM actions. It identifies evidence needed when autonomous agents interact with customer records and enterprise workflows.

## Sources

https://www.salesforce.com/news/press-releases/2024/10/29/agentforce-general-availability-announcement/

https://trailhead.salesforce.com/content/learn/modules/agentforce-agent-planning/define-the-agent-guardrails