Playbook is where you build Skills — reusable automations that save you from repetitive, manual work. You build a Skill once, and it handles the task for you automatically, with no coding required.
In Playbook, you can create two types of skills:
Workspace skills run when something happens in your workspace, such as a chat receiving a bad rating or a new ticket being created.
AI Agent skills run during a conversation between your AI Agent and a customer when the customer's message matches a skill's handling.
Workspace skills vs. AI Agent skills
The quickest way to choose is to ask what starts your skill:
If a system event starts it (a chat ends, a ticket is created, a rating comes in, or it's 9 AM on Monday), create a workspace skill.
If a customer's message to your AI Agent starts it (they ask about an order or want a refund), create an AI Agent skill.
Workspace skill | AI Agent skill | |
|---|---|---|
Starts when | A system event fires, such as a chat ending, a ticket being created, or a schedule | A customer's message matches what the skill handles |
How it runs | Reacts to events, in the background or inside a live chat or ticket | Runs inside the conversation, led by your AI Agent |
Asking the customer questions | Can send a predefined message, but can't wait for the answer and act on it | Asks for information, understands the answer, and uses it in the next step |
Typical use | Notify, route, tag, sync, report, escalate | Answer, look up, qualify, resolve, hand off |
Example | Bad chat rating → create an urgent ticket and notify the support lead | Customer reports a problem without an order number -> ask for it, then create a ticket with the issue summary and tag the chat “support”. |
If your skill needs to ask the customer something and then act on the answer, choose an AI Agent skill. If the data you need already exists in the chat, ticket, or customer profile, choose a workspace skill.
The two types also work well together. For example, an AI Agent skill collects the problem details and creates a ticket. A workspace skill then picks up the new ticket, assigns it to the right team, and posts a summary to Slack.
Creating a skill
To create a new skill:
Open Playbook in Text from the main menu.
Choose how you want to create your new skill:
From scratch: Click the + Create skill button.
From a template: Click Browse templates and select one of the available templates.
How to create a skill from scratch
1. Choose the type of skill
Click + Create skill in the upper-right corner, then choose the type of skill you want to create:
For your workspace: runs on any event — tickets, chats, apps, schedule — or on demand.
For your AI Agent: runs inside your AI Agent’s chats.
2. Describe your skill
The Create skill page opens in prompt mode. Describe what you want in plain language, and Playbook builds the skill for you automatically. For example:
Workspace skill: "When a chat gets a bad rating, create an urgent ticket and send a message to our support lead on Slack."
AI Agent skill: "When a customer reports a problem without giving an order number, ask for it, then create a ticket with the issue summary and order number and tag the chat ‘support’."
You can also:
Pick one of the suggested ideas above the text box, or click Show other ideas to see more.
Type / to add a predefined trigger or action to your description.
3. Review the skill
Playbook builds the skill for you. Review what was created, and then choose what to do next:
Accept it: Click Save and activate in the upper-right corner to make the skill live.
Change it: Rewrite your prompt, and Playbook updates the skill.
Fine-tune it in the builder: Click Go to builder in the upper-right corner and continue building your skill there.
You can also click Go to builder straight away, without writing a prompt, if you want to build the skill step by step.
How to build a skill in the builder
In the builder, you set up your skill block by block.
Key components:
Add trigger — a trigger is an event that starts your skill.
Add action — an action is a task the skill performs after it is triggered.
Every skill must have a trigger and at least one action, but you can add multiple actions if needed.
Triggers and actions can be configured for Text-powered products and 3rd-party software — including dedicated actions for Salesforce and Klaviyo. Some products can be selected as both triggers and actions, while some are available only for individual blocks.
See a step-by-step example of a workspace skill built this way in Example: Follow-up for negatively rated chats, further down this page, right before the FAQ section.
AI Agent skills in the builder
AI Agent skills teach your AI Agent how to handle a specific kind of conversation, including what to ask the customer, how to handle their answers, and when to hand off to a human.
An AI Agent can have up to 10 AI Agent skills.
What the skill knows about the customer
When an AI Agent skill is triggered, it has access to customer context, so the AI Agent can personalize its actions. This includes the customer's:
name,
email,
phone number,
location,
time zone,
the last message,
the chat history,
the page the customer is viewing,
their last five visited pages,
ticket and order counts,
dates of their last chat and ticket.
Creating a skill from a template
You can speed up the process by using pre-built templates:
Go to Playbook and click Browse templates.
Select one of the available templates.
Templates are ready-made workspace skills that open directly in the builder, with the trigger and action blocks already pre-configured. You can customize them to suit your needs.
Runs
The Runs section lets you view your skill logs and check which ones were successful and which ones failed. By clicking on a specific record, you can see the exact step where the skill failed and identify any incorrect configuration. This makes it easier to debug and adjust your skill setup.
You can filter Runs by Status:
All — shows every run.
Completed — runs that were successful.
Failed — runs that encountered an issue with either the trigger or one of the actions.
In Progress — runs that have not yet been completed.
Additionally, you can filter Runs by Time:
All time, Last 7 days, or Last 30 days.
Good to know
Your AI Agent also comes with default skills, such as sending a welcome message or transferring a chat to a human. You can turn them on or off in your AI Agent's settings. To learn more, see AI Agents.
If you only need simple, ticket-only automation — for example, tagging or assigning tickets when a keyword appears — without building a full skill, Ticket rules may be a lighter-weight option.
For conditional logic, variables, and troubleshooting advanced skills, see Advanced Playbook usage.
Example: Follow-up for negatively rated chats
Create a skill that follows up with customers by sending the chat transcript and a personalized message (for example, a request for additional feedback) via a ticket whenever a chat receives a negative rating. The skill creates tickets for your support team, enabling them to review customer feedback, address concerns, and resolve issues more efficiently.
This timely outreach helps you turn negative experiences into recovery opportunities, re-engage customers, and reduce churn. You can further optimize the experience by specifying the time delay between the rating and the delivery of the follow-up message to the customer.
1. Set the trigger
Since the follow-up message should be sent only when a chat receives a bad rating, set the trigger as follows:
Select Chats → Chat rated
Configure this action:
Select Bad rating from the list.
2. Add the first action: Wait
Select Built-in actions → Wait
Configure this action:
Specify the time delay between the rating and the delivery of the follow-up message to the customer.
Duration must be in the format XhYmZs, for example, 1h, 10m, 30s, 1h2m3s.
3. Add the second action: Get chat transcript
Select Chats → Get chat transcript
4. Add the third action: Create a ticket
Select Tickets → Create a ticket
Fill in the required fields:
Subject: Specify the ticket subject.
Message: Click {...} in the upper-right corner of the message field → Chat transcript → Chat transcript parameter. You can also add a custom message here to request additional feedback from the customer.
Requester email: Click {...} on the right-hand side → Chat rated → Chat object → Customer -> Customer email parameter.
Requester name: Click {...} on the right-hand side → Chat rated → Chat object → Customer -> Customer name parameter.
Optionally, you can also configure additional fields such as Status, Priority, Tags, or Followers, though these are not required for the skill to function properly.
5. Publish and test
Once everything is set up:
Go live with your skill by clicking the Publish button.
Test it by starting a chat on your website and verifying that the transcript is automatically sent via ticket.
Frequently asked questions
A workspace skill is triggered when an event occurs in your workspace, such as a chat being rated or a ticket being created. An AI Agent skill starts when a customer's message to your AI Agent matches what the skill handles, and runs inside that conversation.
Yes. For example, an AI Agent skill can create a ticket after collecting the customer's details, and a workspace skill can then route that ticket to the right team.
Yes. You can edit a skill at any time after publishing it. Just remember to publish your changes when you're done so they take effect.