Enable Zendesk Support channel for AI CX Agent to resolve your email support tickets.

Prior to enabling this channel, please make sure there are sufficient agent seats in your account for all of the virtual agents; each virtual agent takes up a full seat with "Staff" permissions.

Integration Instructions

1. Go to Channels and select “Enable” under Zendesk Support. 

2. Enter your Zendesk subdomain. If you have multi-brand organization set up, enter the specific brand's Zendesk subdomain.

3. Authorize app permissions when prompted.

4. Once the connection is completed, you'll end on the "Zendesk Support Integration" settings page.

Settings

• Channels. You can set the desired interception rate for each of the support channels. We recommend keeping the interception rate at We recommend keeping the interception rate at "0" until you've configured your AI Agent, core logics, and other necessary settings.

• Additional Filters. If you have a multi-brand organization set up, you select the brand appropriate brand that's associated with the Shipped Suite account. This ensures that the AI agent will only be intercepting tickets directed to the specified brand.

• External Escalation. Select the brand you would like outside conversation to be escalated to. For example, if a conversation created in the Shipped standalone widget needs to be escalated, the escalation email will be sent to the specified brand. If you have brand filter set up from above, it's generally the same brand for the external escalation.

• Ticket Fields. You can select custom ticket fields that you would like the AI to complete before closing a ticket. If there are certain fields that are only relevant to certain types of tickets, please ensure there's a "N/A" option for the AI to select.

How does it work?

Below describes the flow of how the AI Agent interacts with the customer and what it looks like in Zendesk Support platform.

1. Upon receiving first inbound message, the AI will run an analysis to determine the use case. If it’s been classified as a use case that has been enabled for the AI to handled, it will intercept. If it's not a use case enabled for AI to handle, the ticket will flow into the normal assignment queue.

2. The ticket will be tagged with  shipped-ai-intercepted, indicating that the AI is actively in a conversation with customer. With each response, the ticket will be set to "Pending" status. An AI agent will be assigned to the ticket.

3. The ticket will be marked as "Solved" and tagged with shipped-ai-resolved when the conversation ends. A conversation is ended in two ways:

   1. When the AI has resolved the customer's concern, such as customer has indicated that they don’t need any additional help.

   2. When the conversation has no activity for 72 hours (default idle time).

4. In the case where the AI Agent is unable to assist the customer or the matter requires escalation, the ticket will be tagged with shipped-ai-escalated and moved to "Open" status. The AI Agent is unassigned from the ticket.

5. If at any point, a conversation needs to be manually taken over by a human agent, first add a shipped-ai-intervene tag before responding to the customer. This will ensure that the AI will no longer reply to the customer.

Best Practices

• We recommend minimizing human agent intervention to allow the AI Agent run its course. This is the best way for us to collect data for any configuration changes.

• Create a filtered view of tickets handled by the AI Agent to make it easier for monitoring and QA.

Shipped Related Tags

• shipped-ai-intercepted: The AI Agent is actively in a conversation with the customer.

• shipped-ai-resolved: The AI Agent has resolved the ticket.

• shipped-ai-intervened: A non-Shipped AI Agent has taken over the ticket. The AI Agent will no longer respond to the customer.

• shipped-ai-escalated: The ticket has been escalated to a human agent. The AI Agent will no longer respond to the customer.