> Source: https://builder.ema.ai/v2/templates/customer-support-ai-employee
> Title: Customer Support AI Employee

# Customer Support AI Employee

The Customer Support AI Employee is built from the **Customer Support Chat** curated template (category: Support). It is a chat AI Employee (AIE) that holds multi-turn conversations with your customers: it classifies each incoming message, searches your Knowledge base for relevant articles, and generates a cited response. It reduces first-response time by answering common questions automatically and frees human agents for the issues that need judgment.

## Workflow

The template ships a three-agent workflow. After cloning, the workflow runs left to right:

Node

Agent type

What it does

Classify Intent

`intent_classification`

Sorts the customer's message into one of your support categories. Ships with four: `billing`, `technical`, `account`, `general`.

Search Knowledge Base

`search_respond`

Searches your connected Knowledge bases for articles relevant to the question and returns cited passages.

Generate Response

`respond`

Composes a friendly, cited reply from the search results. If the results are insufficient, it acknowledges the limitation and offers to escalate.

A `start` node feeds the workflow input into the graph, and a `publish` node exposes the final answer. The edges run `start → classify → search → respond → publish`.

The Classify Intent node ships with its intent categories pre-filled in the agent's `type_config`. For example:

```json
{
  "intents": [
    { "label": "billing",   "description": "Billing, invoices, charges, payments, or subscription issues",
      "examples": ["I was charged twice", "How do I update my payment method?"] },
    { "label": "technical", "description": "Product bugs, errors, feature questions, or how-to help",
      "examples": ["The app keeps crashing", "How do I export my data?"] },
    { "label": "account",   "description": "Account access, profile changes, or security concerns",
      "examples": ["I forgot my password", "How do I change my email?"] },
    { "label": "general",   "description": "General inquiries that do not fit other categories",
      "examples": ["What are your business hours?", "Do you have a referral program?"] }
  ]
}
```

## Starting configuration

The template clones in with this configuration, which you can change in the AI Employee builder:

-   **Model selection** — Let Fusion Pick (EmaFusion™ chooses the model per request).
-   **Optimization priority** — Balanced.
-   **Conversation context** — enabled, so the AI Employee retains context across turns.
-   **Feedback** — enabled, with thumbs-up / thumbs-down reasons and free-text comments.
-   **Welcome message** — "Hi! I'm your customer support assistant. How can I help you today?"
-   **Conversation starters** — three buttons: "Report an issue", "Check order status", "Billing question".

**Recommended data sources:** connect your product documentation, FAQ articles, and support policy documents as Knowledge bases so the agent can search for accurate answers.

## How to use

After you select **Use template** and the AI Employee opens in the builder:

1.  **Connect your Knowledge base.** Add your support content — FAQ articles, documentation, troubleshooting guides — to the Search Knowledge Base node. You can upload files or point Ema at web URLs to crawl. See [Knowledge bases](/builder/v2/core-concepts/knowledge-bases).
2.  **Customize the intent categories.** Edit the categories in the Classify Intent node to match your support taxonomy, and update each category's description and examples.
3.  **Adjust the response tone.** Edit the Generate Response node's Instructions to match your brand voice and replace any placeholder company name with your own.
4.  **Publish and deploy.** Publish the workflow, then embed the chat in your support portal or connect it to your ticketing system. See [Channels and deployment](/builder/v2/integrations-data).

> [TIP]
> Test the full conversation before going live. Ask a question you know is answered in your Knowledge base and confirm the reply cites the right source; then ask a question that is _not_ covered and confirm the AI Employee acknowledges the gap and offers to escalate rather than guessing.

## Example queries

-   "I received a damaged product."
-   "What is your refund policy?"
-   "I want to cancel an order."
-   "How do I reset my password?"

## Related

-   [Agent reference](/builder/v2/agent-reference) — the `intent_classification`, `search_respond`, and `respond` agent types in detail.
-   [Knowledge bases](/builder/v2/core-concepts/knowledge-bases) — connecting and crawling your support content.
-   [Templates overview](/builder/v2/templates) — the full curated template list and the clone flow.