Run an action in a connected system. Get a current answer or complete an update. Collect and validate missing details; confirm consequential changes with the caller.
For bookings, order lookups and record changes.
How it works +
Define the authorized operation, input fields, and expected response structure.
Collect and validate missing details; confirm consequential changes with the caller.
Read the returned result before announcing success, and route timeouts or failures to a clear fallback.
IVR navigation
Navigate menus using keypad inputs. Reach the intended department. Send keypad inputs at the appropriate turn and listen for the next prompt.
For calls to automated receptionists.
How it works +
Identify the menu prompts and digits needed to reach the destination.
Send keypad inputs at the appropriate turn and listen for the next prompt.
Handle repeated menus and unsuccessful attempts with a bounded retry or handoff.
Hold & wait handling
Keep callers informed during a pause. Make delays easier to understand. Resume the conversation when the result arrives, preserving the caller’s context.
For slower searches or tool responses.
How it works +
Acknowledge the request before a tool lookup or longer operation begins.
Resume the conversation when the result arrives, preserving the caller’s context.
Define a timeout and a useful fallback when the connected service does not respond.
Voicemail handling
Follow a separate voicemail path. Avoid treating a mailbox as a live contact. Use a concise message that avoids sensitive details when a mailbox is reached.
For outbound reminders and follow-ups.
How it works +
Choose the permitted voicemail behavior for the campaign or workflow.
Use a concise message that avoids sensitive details when a mailbox is reached.
Record the outcome separately and follow the approved retry or follow-up policy.
Specialist handoffs
Move a task to a focused agent. Keep complex journeys organized. Pass the collected identifiers and relevant context needed to continue the task.
For requests that cross responsibilities.
How it works +
Define which intent or condition routes the task to a specialist agent.
Pass the collected identifiers and relevant context needed to continue the task.
Test failed routing, repeated transfers, and the route back to human support.
Human escalation
Bring in a person when needed. Give callers a clear path beyond automation. Configure the receiving destination and the context to share during transfer.
For explicit requests or difficult exceptions.
How it works +
Set triggers for an explicit request, an exception, or an unresolved task.
Configure the receiving destination and the context to share during transfer.
Verify busy and unavailable destinations and offer a clear alternative when no one answers.
Call endings
Close at a defined result or exit point. Separate completed work from unresolved calls. Confirm completed actions only after the connected system responds.
For every completed or interrupted workflow.
How it works +
Define the successful outcome and the conditions for an unresolved exit.
Confirm completed actions only after the connected system responds.
Record the final status and any follow-up required before ending the conversation.
A useful action. In the same conversation.
Scheduling
Check a time, confirm the choice and update the booking.
Order support
Retrieve the record and explain the returned status.
Guest services
Handle an approved request or route to the right team.
Set up the workflow. Make every step count.
Choose the tools needed for this task. Define their inputs, permissions and confirmation requirements.
Can the agent verify information and complete an action during a call?+
Scope the required lookup and action against your connected tools. Define required fields and caller confirmation, then test that the agent reports success only when the tool response confirms it.
How do we handle a caller changing their mind mid-task?+
Identify which steps can be safely cancelled or revised. Reconfirm the new request before an irreversible action, and check the system state if an earlier request may already have completed.
What information should be passed to a human agent?+
Agree on a concise handoff containing the reason for the call, verified details, actions already attempted, and the unresolved issue. Confirm what your destination system can receive and test the receiving team’s experience.
How should failed or slow tool calls affect the conversation?+
Set a clear timeout and fallback for each action. The agent should avoid repeated submissions, explain that completion is unconfirmed, and offer the approved next step.
How do we control high-impact actions such as refunds?+
Use scoped permissions, explicit eligibility rules, and confirmation or staff approval where required. Test unauthorized requests and policy exceptions before enabling an action in production.
