Playbook Anatomy

Every playbook consists of five key components:

1. Steps (Numbered Instructions)

Steps are the building blocks of your playbook. Each step is numbered and describes what happens at that point in the conversation:
Steps execute sequentially by default unless you add branching logic.

2. Tools (Actions the Agent Takes)

Tools are actions prefixed with @ that tell the agent what to do. Examples:

3. Triggers (Activation Conditions)

The trigger defines when your playbook activates:
  • Specific phrase: “I want to book a demo” or “Contact sales”
  • Always active: Leave blank to make the playbook available anytime
  • URL targeting: Use conversation starters to trigger on specific pages

4. Rules (Optional)

Rules let you add specific guidance for how the agent should behave during the playbook. Most playbooks don’t need rules, use them only when you need to add context the agent wouldn’t otherwise have.

5. Branching Logic

Branching lets you create different paths based on user responses:

Structural Rules

How Steps Work

Sequential by Default: Without branching logic, steps proceed to the next number automatically:
One Action at a Time: Each step should focus on a single conversational moment. Turn-Taking: After each step, the user gets a chance to respond before the next step executes.

When to Add Branching Bullets

Add branching bullets when different options lead to different steps:
Omit branching bullets when all options proceed to the same next step:

The Role of @stop_playbook

Tool: @stop_playbook ends playbook execution and returns the agent to general conversation mode. Critical: Every path through your playbook must reach @stop_playbook eventually.

Combining Tools

Most tools can be combined in a single step. A few tools like @initialize_live_chat work best on their own step

Language Handling

  • Tool tags: Must always be in English (@ask_question, @show_options, etc.)
  • Playbook content: Can be written in any language
  • Agent responses: Automatically translated to the visitor’s detected language

Branching Logic Deep Dive

Rule 1: Different Options → Different Steps = Add Branches

Rule 2: All Options → Same Step = Omit Branches

All three products lead to Step 2, so no branching bullets are needed.

Examples: Proper vs Improper Branching

Improper (Unnecessary):
Proper:

Learn More