# AI Agent

The AI Agent handles your frontline customer support autonomously. It reads incoming tickets, finds the answer from your knowledge, and responds — just like a human agent would, but available 24/7.

## How it works

```
 ┌──────────┐      ┌──────────────┐     ┌───────────────┐     ┌──────────────┐
 │ Customer  │────▶│   Trigger    │────▶│   AI Agent    │────▶│   Response   │
 │ creates   │     │  activates   │     │  reads ticket │     │  sent to     │
 │ ticket    │     │  agent       │     │  finds answer │     │  customer    │
 └──────────┘      └──────────────┘     │  takes action │     └──────────────┘
                                        └───────────────┘
                                               │
                                               ▼
                                        Can't answer?
                                        Escalates to
                                        human agent
```

1. A customer creates a ticket or sends a message
2. A **trigger** activates your agent (e.g., "Any Customer Message")
3. The agent reads the ticket and searches your knowledge base
4. If it finds a confident answer, it replies to the customer
5. If it can't answer, it escalates to your team (assign, tag, or internal note)

## Who is it for?

* Support teams handling high volumes of repetitive tickets
* Companies wanting 24/7 coverage without hiring overnight staff
* Teams looking to free up human agents for complex issues

## What it can do

| Capability               | Description                                              |
| ------------------------ | -------------------------------------------------------- |
| **Reply to customers**   | Send public replies or draft responses for review        |
| **Tag and categorize**   | Automatically label tickets by topic                     |
| **Assign tickets**       | Route to the right team member or group                  |
| **Close tickets**        | Resolve simple issues end-to-end                         |
| **Create tickets**       | Open new tickets when needed                             |
| **Leave internal notes** | Share context with your team                             |
| **Look up information**  | Search past tickets, check user details, pull order info |

## Setting it up

1. **Create a Help Desk Agent** from your dashboard
2. **Connect your helpdesk** — [Zendesk](https://docs.eesel.ai/connect-your-tools/helpdesks/zendesk), [Freshdesk](https://docs.eesel.ai/connect-your-tools/helpdesks/freshdesk), [Intercom](https://docs.eesel.ai/connect-your-tools/helpdesks/intercom), [Gorgias](https://docs.eesel.ai/connect-your-tools/helpdesks/gorgias), or [Jira](https://docs.eesel.ai/connect-your-tools/helpdesks/jira)
3. **Train on your knowledge** — Help center, past tickets, documents
4. **Configure actions** — What should the agent be able to do?
5. **Set a trigger** — When should it activate?
6. **Enable Human in the Loop** (recommended at first) — Review replies before they send
7. **Go live** — Turn on the trigger and start handling tickets

For step-by-step instructions, see the [Helpdesk Agent Quickstart](https://docs.eesel.ai/getting-started/quickstart-helpdesk).

## Tips for success

**Start with draft replies.** Configure the agent to leave draft replies instead of sending directly. Review them for a few days, then switch to public replies once you're confident.

**Train on past tickets.** Your agent sounds more like your team when it learns from your actual replies — not just help center articles.

**Write clear escalation rules.** In your [instructions](https://docs.eesel.ai/configure-your-agent/instructions), define exactly when the agent should escalate. For example: "If the customer mentions legal action, immediately assign to the Legal team."

**Use Human in the Loop strategically.** Turn it on for sensitive actions (like refunds) and off for simple ones (like tagging).

## Example scenarios

**Simple Q\&A:** Customer asks "What's your return policy?" → Agent finds the answer in your help center and replies.

**Order lookup:** Customer asks "Where's my order?" → Agent uses the Shopify integration to check order status and replies with tracking info.

**Escalation:** Customer reports a bug the agent can't resolve → Agent leaves an internal note summarizing the issue and assigns to the Engineering team.

## Related

* [AI Copilot](https://docs.eesel.ai/use-cases/ai-copilot) — If you want AI to draft replies for your team instead of sending directly
* [AI Triage](https://docs.eesel.ai/use-cases/ai-triage) — If you want to auto-categorize tickets alongside your agent
* [Actions reference](https://docs.eesel.ai/configure-your-agent/actions) — Full list of available actions
* [Triggers reference](https://docs.eesel.ai/configure-your-agent/triggers) — All trigger options