> ## Documentation Index
> Fetch the complete documentation index at: https://docs.theymes.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Assignment strategies

> Choose when AI agents receive tickets through manual actions, workflows, and automation triggers.

Assignment controls when an AI agent takes responsibility for a ticket. The agent must be **Ready**, the ticket must be open, and its app must be enabled for the agent to handle it.

Creating instructions or adding skills does not route tickets. Choose an assignment strategy that matches how much control your team needs.

| Strategy | Who decides | Use it for |
| - | - | - |
| Manual assignment | A team member selects the agent on a ticket or through a bulk action. | Trying the agent, reviewing suitability, or returning a ticket after human work. |
| Manual workflow | A team member runs a published workflow that assigns the ticket. | A repeatable handoff with filters or other ticket updates. |
| Automation trigger | A published automation assigns matching tickets when an event happens. | Routing a defined category of incoming support automatically. |

## Manual assignment

1. Open an eligible ticket in the agent desktop.
2. Open its assignee menu.
3. Select the AI agent by name.
4. Review its response and subsequent ticket outcome.

You can also assign tickets through bulk actions. Use a selection whose apps are enabled for the agent.

For manual assignment, leave **Hide from assignee menu** off in the agent's setup. When on, the agent cannot be selected through ticket or bulk assignment actions.

The agent uses the existing conversation to understand the request. It continues handling customer replies and collector submissions while the ticket remains assigned to it.

### Take over or return a ticket

Assign the ticket to a human when your team needs to take over. The AI agent stops handling it once it is no longer the assignee.

A human can return a ticket to the AI agent after handling the part that needed human support. The agent continues from the latest conversation and can wait if the customer still needs to answer a question.

Before returning a ticket, explain any completed actions or next steps in a message to the customer. The AI agent uses those messages to continue; it does not read private notes.

## Assign through a manual workflow

Create a [manual workflow](/product/automation/workflows.md) when team members should start a standard handoff rather than selecting the agent directly.

For example, a workflow named `Ask app support agent` can:

1. Use a **Manual workflow** trigger filtered to the relevant apps and ticket categories.
2. Add a tag identifying tickets sent to the agent.
3. Run **Assign ticket** with the AI agent selected as the assignee.

Save, publish, and enable the workflow. Team members can run it from tickets where its scope and filters match. Workflow assignment works even when **Hide from assignee menu** is on.

## Assign through automation triggers

Use a [trigger automation](/product/automation/triggers.md) to assign tickets automatically. A straightforward starting point is **Ticket created** for a narrow category of questions covered by your knowledge base.

1. Create a trigger automation and scope it to apps enabled for the agent.
2. Add a **Ticket created** trigger.
3. Add filters for the tickets it should handle, such as a support category, tag, language, or custom field. Include an assignee filter when the rule should only apply to unassigned tickets.
4. Connect an **Assign ticket** action.
5. Set **Ticket ID** to the ticket ID from the trigger, typically `{{ trigger.ticketId }}`.
6. Set **Mode** to **Static** and select the AI agent under **Assignee**.
7. Save, publish, and enable the automation.
8. Create a matching test ticket and inspect both the automation run and the agent's conversation.

See [Actions and switches](/product/automation/actions-and-switches.md#ticket-assignment-status) for **Match status to assignment**. Assignment does not reopen a closed ticket or remove an existing hold.

### Route different categories to different agents

Use **AI Tags & Fields** to categorize tickets based on what the customer needs. Then use those tags or fields in your automation's trigger filters to route each category to the appropriate specialist agent with **Assign ticket**.

For example:

* App usage questions → app support AI agent.
* Login troubleshooting → login AI agent.
* Payment disputes or account ownership requests → human support.

See [trigger filters](/product/automation/triggers.md#trigger-filters) for routing tickets by tags and fields.

### Assign on later activity

You can also assign tickets through **Customer message**, **Ticket events**, or **Ticket waiting** triggers. An already assigned AI agent handles new customer replies automatically.

Avoid broad rules that reassign every customer message to AI. They can take tickets back from humans or return escalated tickets before human work is complete. Filter on the intended assignee, category, and status, and check other routing automations for competing assignments.

## Plan human handoff

When an AI agent escalates, it tells the customer a human will take over and **unassigns the ticket**. It does not choose a particular human or guarantee an immediate response.

Make sure your human routing or triage process covers these tickets. Check that your AI assignment automations do not immediately send them back to an AI agent. A scoped **Ticket created** rule is a useful starting point because it does not run again merely because the agent escalated.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.