If your customer base uses multiple languages for communication, there are several settings you can take advantage of to improve their experience during conversations with your AI agent.
This article walks you through several recommended steps to take when supporting a multilingual customer base through your AI agent.
This article contains the following topics:
- Translating AI agent messages
- Translating help center articles
- Translating conversations in the Agent Workspace
- Testing multilingual AI agents
Related article:
Translating AI agent messages
The best place to start when supporting a multilingual customer base is to make sure your AI agent’s messages are translated into the languages your customers use.
The easiest way is to configure your AI agent to automatically translate its messages. This includes its standard responses and (for legacy functionality) answer steps. See Activating automatic translation for an AI agent.
If you find that the automatic translations for some AI agent messages don’t meet your standards—for example, by translating product and brand names, using incorrect pronouns, or using inaccurate phrasing—you can replace them with custom translations. See Creating custom translations for AI agent messages.
Translating help center articles
After translating your AI agent messages, the next best step is to make sure that your help center content is also available in the languages your customers use.
AI-generated replies sent by the AI agent might suggest help center articles in a conversation to direct customers to information that may help them self-solve their questions. To offer help center articles in multiple languages, you’ll need to activate those languages, then add translated versions of those articles to your help center.
For more information, see Configuring your help center to support multiple languages and Localizing help center content.
Translating conversations in the Agent Workspace
If your human agents work in different languages than your AI agent, they might need to translate the content of a conversation between a customer and your AI agent if the conversation is escalated.
After taking over a conversation from the AI agent, human agents can use the native translation feature in the Zendesk Agent Workspace to communicate with customers and other agents across multiple languages. This feature is activated by default for all customers using Agent Workspace, but is available only for agents with a Chat seat.
For more information, see Translating conversations in the Zendesk Agent Workspace.
Testing multilingual AI agents
Finally, you should always test your AI agent in the languages your customers will use it in. This helps you evaluate the automatic and custom translation results and understand the end-user experience. See Testing an AI agent before publishing it for customers.