Control when callers can interrupt your AI Agent

Some callers talk over the AI Agent. Some AI Agents need to finish a sentence before listening. Both are right, depending on the moment:

• A caller who knows what they want shouldn’t sit through a full greeting. Let them cut in.
• A compliance disclosure, a cancellation notice, or a “this call is recorded” statement has to be heard in full. The AI Agent finishes, then listens.
• “Is this for your home or business account?” fails if the caller answers after “home” and never hears “or business.” The question has to land completely before the AI Agent takes an answer.

Now you decide which behavior applies, at every level: agent-wide, per block type, or for a single block in a process.

What’s new

Channels → Voice → Interruptions, a new tab.

Set the agent-wide rule: On lets callers interrupt, Off makes the AI Agent finish speaking first. Below it, override the rule per block type: Message, Collect Data, Options, and External Agent.

Staging and Live are configured separately, so you can try a change on Staging before any caller hears it.

Every setting is Default, On, or Off, and Default shows the built-in value in brackets: On for Collect Data, Options, and External Agent; Off for Message.

Allow interrupting, on every voice block.

Each block now has a Default / On / Off select. Default inherits from Channels → Voice → Interruptions.

Your existing settings are respected. Any block with Speak message in full turned On is now set to Off for interruptions, in both Staging and Live. Nothing to redo.

How the settings combine

If the agent-wide setting is Off, nothing can be interrupted and block settings are ignored. The UI tells you so. If it’s On, the block’s own setting wins, then the block-type override.

Good to know

• Changes take effect from the next conversation turn. No publish needed.
• Re-ask the question when its key part was not heard is now Re-ask if interrupted. Same behavior, shorter name.
• Viewing the tab requires process access; editing requires the Process update permission. Every change is written to the audit log.
• Available for AI Agents migrated to optimisied Zowie architecture.

In case of questions, please contact your Technical Account Manager.