> ## Documentation Index
> Fetch the complete documentation index at: https://docs.stardeck.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Playbooks
> Give a customer agent a goal to work toward in conversation, then a set of ways to hand the customer to a human when it's done.
Where [guidance cards](/customer-agents/guidance-cards) are a library of unordered rules, a **playbook** is a single goal-directed flow. You give the agent a goal, an ordered set of things to cover, and the ways it can route the customer to a human—and it works toward that goal naturally across the conversation.
A playbook is **optional** and there's **one per agent**. A good fit is qualifying an inbound lead, booking a consult, or triaging a request before a person takes over.
## What a playbook contains
* **Goal** — one line describing what the agent is trying to achieve. *"Qualify an inbound lead and route them to sales."*
* **Objectives** — an ordered list of things to cover. The agent weaves them into the conversation rather than interrogating the customer—it won't fire them off as a checklist.
* **Pacing note** *(optional)* — guidance on timing. *"Have a couple of exchanges before offering to hand off."*
* **Handoff options** — the ways the agent can offer to continue with a person (see below).
```mermaid theme={null}
flowchart LR
Start["Customer arrives"] --> Work["Work through objectives\nin natural conversation"]
Work --> Done{"Objectives covered\nor customer asks?"}
Done -- "yes" --> Offer["Offer a handoff option"]
Offer --> Pick["Customer picks one"]
Pick --> HO["Hand off to a human →\nagent stops replying"]
```
## Handoff options
Handoff options are the menu the agent offers once the goal is met (or the moment the customer asks for a person). Each option has a **type** and a **label**—the label is what the customer sees:
| Type | Example label |
| -------------------- | -------------------------- |
| **Phone** | "A phone call" |
| **Online meeting** | "A video call" |
| **On-site meeting** | "An in-person meeting" |
| **LINE salesperson** | "Chat with our sales team" |
| **Operator** | "A member of our team" |
When the customer picks one, the agent hands the conversation to a human and **stops replying**. An **operator** handoff is always available as a fallback, even if you don't add it explicitly—the agent can always escalate to a person.
A handoff option is an *offer and a routing label*, not an integration. Choosing "A phone call"
marks the conversation as handed off and tells your team how the customer wants to continue—your
team takes it from there.
## What happens at handoff
Once the agent hands off:
* The conversation is **marked as handed off** and the agent won't reply again—on any channel—until your team resumes it.
* It surfaces in the [Conversation Pool](/customer-agents/operator-workflow) for a person to pick up, with the route the customer chose.
* Reply mode is independent of handoff. A handoff stops the agent regardless of whether it was in auto-send or auto-draft.
See [handing off to a human](/customer-agents/operator-workflow#handing-off-to-a-human) for the operator side and how to resume.
## Playbook vs. guidance cards
| Use a… | When you want… |
| ----------------- | ----------------------------------------------------------------------- |
| **Guidance card** | A standing rule that's always true—tone, formatting, what to avoid. |
| **Playbook** | A goal with a beginning and an end—gather what you need, then hand off. |
They compose: an agent can run a playbook *and* follow its conventions at the same time.
## Next steps
Pick up a handoff and resume the agent
The conventions that run alongside a playbook
Set a playbook under Behavior