Release notes

Playbooks now give you more control over how your AI Agent handles every part of a conversation — with faster execution and more predictable outcomes.

What’s changing

Playbooks are now built around a structured step-based model that lets you define exactly how your Agent executes each phase of a conversation. The legacy freeform authoring format is now called Playbooks (Classic).

More control over each step

Every Playbook is built from six discrete step types, each performing a single, well-defined operation:

• SEND: Send a message to the end user (Fixed message or Contextual AI-generated).
• SET: Assign a variable value silently (Exact value or Use reasoning mode). Never prompts the end user.
• ASK: Ask the end user for a value (Capture exact response or Map using reasoning mode).
• RUN: Execute an Action, Handoff, linked Playbook, or Exit.
• IF/ELSE: Branch the flow based on conditions, with support for nested AND/OR condition groups.
• GO TO: Jump to a specific step to enable loops, retries, and non-linear flows.

For each step, you choose the mode — fixed or AI-driven — so you decide where the Agent follows your instructions exactly and where it reasons on its own.

Built-in flow intelligence

Playbooks now monitor conversations during execution and intervene automatically:

• Intent change detection: Exits the Playbook and reroutes when the end user’s intent shifts.
• Pause and resume for Knowledge search: Pauses execution to perform a Knowledge search, then resumes where it left off.
• Automatic escalation: Escalates to a human agent or exits the Playbook based on end user input, even without an explicit Handoff or Exit step at that point in the flow.

Action error handling

When an Action returns a 4xx or 5xx HTTP status code, save the status code to a variable and branch on it with IF/ELSE to build custom fallback behavior.

Multi-channel support

Playbooks are supported on Messaging, Email, and Voice channels.

Learn more in the Playbooks documentation.