The “agentic call center” is the next big step in customer service, transforming rigid, rule-based IVR chatbots into resilient and adaptable agentic teammates. But like any new hire, their success depends on how well they’re integrated with your existing human workforce. For a long time, the handoff from a voice bot to a person was a bit of a black box—it usually happened automatically the moment the bot reached its limit and simply gave up. But in the agentic enterprise where humans and agents are working together, it’s more than just what an agent can do on its own; it’s about knowing exactly how and when to pass the baton to a human.
With Agentforce, we’re moving away from these “emergency bailouts” toward a smarter, more organized way to handle calls. By using the patterns we’ll cover here, you can turn your Voice Agent into a high-level gatekeeper. This ensures your human team only steps in when they are truly needed, and only after the AI has finished essential pre-work—like verifying an account or opening a support case. This transforms the handoff from a “fail-over” into a strategic advantage that saves time for everyone.
1. Default Escalation Behavior: Out-of-the-Box
By default, escalation is treated as a standard, automatic action triggered at the topic level. When a user explicitly asks to speak to a human, the system bypasses internal logic to fire a handoff immediately.
| Component | Description |
| Trigger | User request (e.g., “I need to talk to a human.”) |
| Mechanism | The Topic classifier identifies the “Escalation” topic. |
| Action | A hidden “Escalation Action” is executed, firing the Omni Flow configured in the system’s Escalation section. |
| Limitation | Lack of control. The classifier bypasses any instructions within the topic, preventing the agent from collecting data or running tools before the handoff. |
2. Dynamic Escalation: Contextual and Controlled
Dynamic Escalation allows for specific logic and “if-then” policies. This ensures that automated steps are exhausted first and that the human agent receives a call that is already contextualized and verified.
How to Implement
To enable this level of control, you must authorize the specific topic to handle escalations via a metadata update:
- Set the Metadata Flag: Use the Developer Console, Workbench, or an IDE to query the GenAiPluginDefinition object.
- Update the Record: Locate the relevant topic record and set the canEscalate field to True.
- Use the “Escalation” Keyword: When writing your topic instructions, you must use the keyword “Escalation” (in quotes), as this is the specific backend action name the system recognizes.
Implementation Patterns and Examples
Once the flag is set, you can use instructions to enforce business rules. Note the specific wording required for the agent to recognize the action:
- Pattern A: Information Collection (e.g., Email Verification)
- Goal: Collect the user’s email before escalating.
- Instruction: “If the user asks to talk to a human agent, allow the ‘Escalation’ action only if the user has provided their email address. If they have not, ask for the email first.”
- Pattern B: Required Action Execution (e.g., Case Creation)
- Goal: Ensure a case exists before the handoff.
- Instruction: “…allowing the ‘Escalation’ action to be called only after the ‘createcase’ action has been executed in the Case Management topic.”
3. The “Smart Routing” Strategy
In scenarios where the Voice Agent’s primary purpose is to act as a high-fidelity dispatcher, we recommend the Smart Routing pattern. This setup focuses the agent’s intelligence solely on capturing intent and routing the call.
- Setup: Implement a single “Escalation” topic.
- Classification Description: “Handle all escalations and any user requests that come through.”
- Logic: Under this generic topic, include one simple instruction to escalate the conversation to a human live agent.
Moving from automatic to dynamic escalation turns the handoff from a “bailout” into a strategic advantage. By implementing the canEscalate flag and using the “Escalation” keyword in your instructions, you empower your Voice Agent to be a more effective gatekeeper. This results in reduced agent transfers, higher-quality data for your live agents, and a more seamless experience for your customers.
To learn more about Agentforce Voice, visit here.
