Zendesk Chat Handoff usage

Overview

By integrating Zendesk Chat and Ada, you can provide end users with a frictionless experience that instantly hands them off from the AI Agent to a live agent during a Handoff. End users begin their conversation with the AI Agent, and if required, live agents are invited to join the chat—all within the same interface.

This article outlines the experience on the Zendesk side once you have configured Ada with your Zendesk environment. For detailed instructions on configuration, see Configure Zendesk Chat.

Limitations

The following limitations apply to Zendesk Chat Handoffs:

• Live agent surveys are not supported.
• Rating requests from agents are not supported.
• Zendesk triggers are not supported.

Capabilities & configuration

Zendesk Chat Handoffs provide a seamless transition between your AI Agent and live agents. This section describes available features, as well as the experience for end users and agents.

Transcript summaries

Ada can automatically attach a summary of an end user’s conversation to a Handoff transcript. This summary helps human support agents quickly determine the reason for the escalation.

• The feature examines the end user’s messages and looks for key phrases that represent what the end user is trying to ask or do.
• The summary includes up to three end user messages.
• The summary appears after the last message of the transcript.
• Summaries only consider the end user’s text messages, not button clicks or AI Agent messages.
• Summaries are available for conversations in all languages.

File transfer

File transfer in Zendesk Chat provides two-way file sharing, meaning both end users and agents can share files within the live agent chat conversation. By default, end users and agents can share the following file types: JPEG, PNG, GIF, PDF, and TXT, to a maximum of 20 MB, with only one file being shared at a time. To enable additional filetypes, you can add them in your Zendesk settings. To learn how, see Managing file sending in live chat at Zendesk’s documentation.

The end user can either drag and drop or click the paperclip icon to select the file from their computer. If they click the paperclip icon, they see a window in which to select their file—only files that meet the limitations mentioned above (file type and size) are selectable, and all other files are greyed out. If the end user drags and drops an unsupported file, they receive an error message alerting them that the file is too large, unsupported, etc.

The experience for the agent is the same as the standard Zendesk live agent chat, where agents can click the file icon in the Zendesk live chat interface to share a file with the end user.

Once shared, the files are hosted in Zendesk, but those with access to the Ada dashboard can view the files from the Convos view.

Supported features

Zendesk Chat Handoffs support the following capabilities:

• For clients on Zendesk Agent Workspace—Tags are supported, as well as the automatic mapping of name, email, and phone variables to ticket fields as long as they are lower case global variables in Ada.

• Agent-to-agent transfers, cold transfers, and transfers to departments are supported in both Zopim and Zendesk Agent Workspace.

• The following transfers are supported for both Zopim and Zendesk Agent Workspace:

  • Agent-to-agent (transfer and leave, or invite)
  • Transfer to department

Customer experience

When an end user reaches the point in the Handoff dialog where the Zendesk Live Agent block has been placed, they are added to the appropriate queue, provided the end user is within your support department’s operating hours. Within their chat window, they see a banner indicating their queue position and can no longer chat with the AI Agent.

The agent joins the conversation in the same chat window, creating a seamless experience. The agent’s name and Ada live agent avatar display in the chat window. To personalize the Ada live agent avatar, navigate to Channels > Chat > Appearance > Chat window > Avatar.

Agent experience

When an end user reaches the point in the Handoff dialog where the Zendesk Live Agent block has been placed, Ada routes the conversation to your live agent(s), placing them in a queue until a human agent becomes available. The Handoff process relies on the Zendesk routing mechanisms, so ensure that your departments or groups and hours of operation are appropriately configured within Zendesk.

The Zendesk department or group that you identify in the Zendesk Live Agent Block is the department end users are routed to for Handoff. If there are no agents available in the preferred department—either due to your queues being full or all agents being offline—the end user is routed to the Off-Hours dialog.

If the request to hand off to a live agent comes outside of your live chat hours of operation (as set in Zendesk), the AI Agent routes the end user to the Off Hours dialog. Your live agent can then choose to accept the chat from the Zendesk dashboard, as normal.

If the AI Agent has captured specific information from the end user, such as their name, email, or phone number, that information maps to the chat, as do any tags you have set up on the Zendesk Live Agent Block. The agent also sees the conversation the end user had with the AI Agent leading up to the Handoff, as well as any tags included in the Live Agent Block.

The transcript handed over from the AI Agent to the live agent includes all messages up to the point when the end user transferred to the live agent queue. Note that the agent cannot see what the end user is typing before they send the message—they see that they are typing, but not what they are typing.

When your live agent is speaking with an end user who transferred over from the AI Agent, the live agent has access to all of their usual Zendesk actions, including transferring chat to another agent, creating a ticket, and more. When an end user speaks with an agent, the chat populates in the ticket as a public comment; however, in email tickets, the chat populates as a private comment.

If an end user goes idle (triggered between 20 seconds and 2 minutes of not having the chat window in focus, or 20 mins of inactivity), the conversation with the agent automatically ends, and the end user is provided with any dialog blocks that follow the Zendesk Live Agent Block. The end user can re-escalate if they would like to reconnect to an agent.

FAQs

Common questions about Zendesk Chat Handoffs.

Q: My conversations appear to be disconnecting, why is this?

There are a few reasons why the conversation may be disconnected:

• If the end user is on a mobile device and the device goes to sleep, the live agent is disconnected (this is intended behaviour from Zendesk and not Ada-related).

• If the end user is using a computer with browser privacy enabled and navigates away from the page, the live agent is disconnected.

Q: Can more than one agent be involved in a conversation?

No. Currently only one agent can be involved in a conversation at a time. If more than one agent is involved, all information is routed to the first original agent. Any subsequent agents do not receive any information.

Q: What happens when the live agent ends a chat?

Any blocks, as well as quick replies, can be included after the live chat Handoff block, so that when live chat ends, the AI Agent can re-greet the end user.

Q: After agent to agent transfer will the second agent get the same transcript?

Yes, they will.

Q: Do you support the Zendesk automatic translation feature?

Yes. Ada is also multi-lingual, so all languages are supported out of the box. For example, messages announcing the end user’s queue position, presence messages, etc. are all translated.

Q: Do you support emojis in live chat?

Yes.

Q: Is Zendesk Chat through Ada supported on mobile?

Yes, on iOS and Android.

Q: I have set everything up, but when I attempt to hand off to an agent, I get a ‘Conversation ended’ message.

The Zendesk API’s client secret or ID have probably expired or were deleted. Re-generate the client secret and update it in the Settings > Platform > Zendesk Live Chat platform modal. If that doesn’t work, you may need to whitelist the AI Agent by adding the AI Agent’s URL to your whitelist.