Agentforce 2dx extended agents beyond the conversational request, allowing them to work proactively from business events and embedded experiences. This removed a major limitation—and removed a natural control point.
When no person initiates the interaction, the trigger, eligibility logic and stopping condition become the core of the product.
Cloud Group point of view
Proactivity should begin with high-confidence signals and reversible work. The agent may decide how to complete a bounded job, but deterministic policy should decide whether the job is allowed to begin.
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:
- Define the triggering event and the evidence required to qualify it.
- Add deduplication, cooldown and maximum-attempt rules.
- Start with drafts, recommendations or reversible record updates.
- Notify an accountable human when thresholds or exceptions are crossed.
- Test event storms, stale data and partially completed actions.
The architecture and operating implication
Use platform events, Flow or APIs to create a controlled work item with an idempotency key. Let the agent operate within that scope, recording state transitions and tool calls. A separate policy service should be able to pause a class of work without redeploying the agent.
Measure what changes
Model activity is not a business result. Track a small set of indicators that connect behavior to accountable work:
- Qualified triggers versus false starts
- Duplicate or repeated actions
- Value created before human intervention
- Time from event to resolved outcome
The value of a proactive agent is not that it acts first. It is that it recognizes the right moment and completes the right bounded work without creating another queue.
Primary sources
This field note is grounded in the product and market context available at the time of publication.
