As Agentforce approached general availability in October 2024, the difference between a demo and a production service became concrete. Production brings real identities, imperfect records, concurrency, policy exceptions and customers who do not follow the script.
Readiness must cover behavior, platform controls, operations and economics. Passing functional tests is necessary, but not sufficient.
Cloud Group point of view
Launch only when the organization can observe, interrupt and improve the agent. Autonomy without an operating system for accountability is unmanaged outsourcing.
A practical playbook
The strongest next step is narrow enough to govern and useful enough to produce evidence. We would structure the work around these moves:
- Approve the job charter, action allowlist and explicit prohibited behavior.
- Validate every action with least-privilege personas and negative tests.
- Baseline evaluations across common, rare and high-consequence scenarios.
- Staff the escalation path and define rollback or kill-switch authority.
- Set consumption alerts and cost expectations before demand arrives.
The architecture and operating implication
Separate environments, version agent configuration and preserve trace evidence. Ensure knowledge freshness has an owner and action failures return safe, intelligible responses. Connect production telemetry to the same backlog and release process used for Salesforce improvements.
Measure what changes
Model activity is not a business result. Track a small set of indicators that connect behavior to accountable work:
- Critical evaluation pass rate
- Successful autonomous completion
- Handoff accuracy and response time
- Cost per outcome and abnormal consumption
A controlled launch does not eliminate uncertainty. It ensures uncertainty becomes visible, bounded and actionable before it reaches the customer at scale.
Primary sources
This field note is grounded in the product and market context available at the time of publication.
