> For the complete documentation index, see [llms.txt](https://docs.klink.cloud/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.klink.cloud/advanced-setup/workflow-automation.md). # Workflow Automation Workflows automatically assign, reply to, tag, score, or route conversations. Each rule follows **When (trigger), If (conditions), Do (actions)**. **Where:** **Workflow** in the left menu (page title "Workflow Automation"). {% hint style="info" %} For on-demand (manual / custom) workflows, including how agents run them from a conversation or in bulk from the CX Log, see [Manual Workflows Shortcut](https://docs.klink.cloud/advanced-setup/setup-workflow-automation/manual-workflows-shortcut). {% endhint %} ## What workflows are A workflow is an automation rule. When a trigger event happens, klink.cloud checks optional conditions, then runs actions in order. Use workflows to: * Route new chats to the right queue or agent * Let Kai (Chat AI Agent) answer first * Send after-hours replies * Score sentiment and CSAT when a conversation closes * Nudge silent customers * Auto-fill case log fields and post AI summary comments for handoffs * Give agents one-click wrap-up shortcuts (manual workflows) * Create tickets from conversations (optionally with AI subject and description) ## The workflow list * **Create:** click **New workflow**. * **Enable / Disable:** the toggle in each row. Disabled workflows do not run. * **Priority / order:** drag a row, or use the **Hierarchy** buttons (Move to top / up / down / bottom). Rules at the top are evaluated first when several match the same conversation, so put the most specific rules highest. * **Duplicate / Delete / Edit:** use the row menu (⋮). Click a workflow's name to edit it. * **Custom** badge (⚡): marks a manual **When run manually** workflow. See [Manual Workflows Shortcut](https://docs.klink.cloud/advanced-setup/setup-workflow-automation/manual-workflows-shortcut). ## Anatomy of a workflow Every workflow is built in three steps. ### Step 1: Trigger (When) Add one or more trigger events. Most combine with **OR** (any one fires the rule); some are exclusive. | Trigger | What it does | Extra setup | | --------------------------------------- | ----------------------------------------------------------------------- | ------------------------------------------ | | **When new chat room is open** | Runs when a customer sends their first message. | none | | **When a chat room is closed** | Runs every time a conversation is closed. | none | | **When a customer sends a message** | Runs every time a customer replies. | none | | **When chat room is on hold** | Runs on rooms held for a period. | Hold timer (hours, minutes). *Exclusive.* | | **When an AI agent transfer chat room** | Runs when an AI agent hands off to a human. | none. *Exclusive.* | | **When an AI agent session times out** | Runs when an AI conversation goes idle past its timeout. | nothing to set. *Exclusive.* | | **When customer didn't response** | Runs if the customer is silent for a period. | Timer (days, hours, minutes). *Exclusive.* | | **When agent didn't response** | Runs if the agent is silent for a period. | Timer (days, hours, minutes). *Exclusive.* | | **When run manually** | Never runs on its own; an agent starts it from the Run workflow button. | none. *This makes it a custom workflow.* | ### Step 2: Conditions (If) Optional. Conditions combine with **AND** / **OR**. Operators include **is / is not**, **contains / does not contain**, **starts with / ends with**, **Empty**, and for business hours **Inside / Outside of business hour**. | Condition | What it filters by | | ---------------------------- | ---------------------------------------------------------------------------------------------------------- | | **AI Conditions** | Detect intent using a plain-English prompt (for example "Customer is asking for a refund or complaining"). | | **AI Conditions (Advanced)** | Analyze the whole chat transcript with a custom prompt, and branch actions by its output. | | **Hold label** | The hold-label status. | | **Contact info** | Name, Contact label, Phone, Mail. | | **Language** | The message language. | | **AI Language** | The AI-detected language. | | **Chat info** | Assignee. | | **Channel info** | Messenger, LINE, Viber, etc. | | **Message content** | Keywords in the message. | | **Business hour** | Inside / Outside a business-hours schedule. | #### How AI Conditions (Advanced) works This is the recommended path for sentiment and CSAT scoring on the full conversation: 1. **Left side:** an **AI Prompt** with **versions** (v1, v2, v3). Write a prompt that returns a single label (for example `very_satisfied / satisfied / neutral / unsatisfied / very_unsatisfied`), and end it with "Reply with only the label". Add versions to experiment; only the active version runs. 2. **Right side:** **Test on chat history**. Click **Select chat**, preview a real transcript, and click **Run prompt** to see the AI's output so you can tune the prompt before saving. 3. When you use an Advanced condition, the Actions step lets you **branch by the AI's output**. {% hint style="warning" %} With the **When run manually** trigger, the **Message content** and **Hold label** conditions are hidden (a manual run has no triggering message or hold status). {% endhint %} ### Step 3: Actions (Do) Actions run in sequence when the trigger and conditions are met. | Action | What it does | | ----------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **AI Agent** | Let a live AI Agent (Kai) handle the conversation. | | **Assign Human Agent** | Route to a specific agent or a **queue**. | | **Send Message** | Send a text reply: **Fixed Message** (you type it) or **AI Message** (AI-generated). | | **Send Rich Message** | Send text plus images, videos, or attachments. | | **Add contact label** | Add labels to the contact. | | **Add Chat Tags** | Add tags to the case. | | **Add Note** | Add a note to the case log. | | **Add Comment** | Post an internal team-only comment (supports @mentions). Use **AI Generated** mode to draft a conversation summary from a prompt. | | **Set Disposition** | Set the case disposition. | | **Set Categories** | Set the case categories (up to 3 levels). | | **Set Type** | Set the case type. | | **Set CSAT Score** | Set a fixed CSAT (Very satisfied to Very unsatisfied). | | **AI CSAT Score** | Let AI read the transcript and score CSAT. | | **Set Sentiment Score** | Set a fixed sentiment (Positive / Neutral / Negative). | | **AI Sentiment Score** | Let AI read the transcript and score sentiment. | | **Auto Case Log** | Let AI read the transcript and fill in **blank** case-log fields. Pair with tags, categories, notes, or comments for fuller enrichment. | | **Create Ticket** | Create a ticket from the conversation, with optional AI subject and description from the transcript. See [Create Ticket](/advanced-setup/workflow-automation/recipe-create-ticket.md). | | **Send Webhook** | Send an HTTP request to an external URL (with variables like message, contact, transcript). | | **Close Chat** | Auto-resolve and close the case (optionally with a disposition). | #### Branching by AI output Only available with **AI Conditions (Advanced)**. Choose **Branch by AI output**, add **Output labels** matching what your prompt returns (for example `very_satisfied`, `very_unsatisfied`), and configure the actions for each label plus an **Else** fallback. Example: `very_satisfied` sets CSAT to Very satisfied; `very_unsatisfied` sets CSAT to Very unsatisfied and adds a comment to alert the team. ## Important rules * **Queue assignment** (Assign Human Agent, then Queues) can only be used with the **When new chat room is open** trigger. * **AI Agent** and **Assign Human Agent** cannot both be in the same workflow. * The **on-hold** trigger only allows **Close Chat**, **Send Message**, **Send Rich Message**, and **Add contact label**. * Most actions can be added once; **Send Message**, **Send Rich Message**, and **Add contact label** can be added multiple times. ## Custom (manual) workflows A **custom workflow** uses the **When run manually** trigger. It never fires on its own. An agent runs it on demand from the **Run workflow** (⚡) button on a conversation, or in bulk from the CX Log. Only **active** custom workflows appear in the Run workflow picker. Full create and run steps: [Manual Workflows Shortcut](https://docs.klink.cloud/advanced-setup/setup-workflow-automation/manual-workflows-shortcut). ## Step-by-step recipes | Recipe | Guide | | -------------------------------------------------- | ---------------------------------------------------------------------------------------------------------------- | | 1. Auto-assign new chats to a queue | Route to a queue | | 2. Let Kai answer first | Kai first response | | 3. Auto-reply outside business hours | After-hours auto-reply | | 4. Score sentiment and CSAT on close (recommended) | Sentiment and CSAT on close | | 5. Chase a silent customer | Silent customer nudge | | 6. One-click wrap-up (custom workflow) | One-click wrap-up | | 7. Auto Case Log (fill blank case-log fields) | [Auto Case Log](/advanced-setup/workflow-automation/recipe-auto-case-log.md) | | 8. AI conversation summary comment | [AI conversation summary comment](/advanced-setup/workflow-automation/recipe-ai-conversation-summary-comment.md) | | 9. Create Ticket from a conversation | [Create Ticket](/advanced-setup/workflow-automation/recipe-create-ticket.md) | Also see: AI CSAT scoring workflow (aligns with Recipe 4; replaces the old multi-workflow pattern).