Language AI allows consumers to engage in their native tongue, even if agents don’t speak the language. Monolingual agents can now provide multilingual support with machine translation built right into the Helpshift platform. With Language AI: Translation of Agent-User Conversations, agents can focus on what they do best—making consumers happy.

Language AI automatically detects the issue language for translations. If needed, the agent can manually change the issue language to any of the supported languages.

If the agent’s default language is English and the consumer submits a query in another language, a Translate button appears next to the end user's message.

Clicking it instantly translates the message into the agent's language (English). Both the original and translated messages are shown in the conversation, making communication smooth and effective.

Language AI works with two types of translation engines:

  • Generic Engines
  • Custom-Trained Engines

By default, Language AI uses pre-trained generic engines for translations. However, Language AI is flexible and can be configured to work with custom-trained engines based on branding requirements.

Generic Engines

Generic engines are pre-trained AI models designed to translate text from one language to another. They're created by training on a diverse set of data sources to ensure broad coverage of languages and content. List of supported languages for generic engines

Custom-trained Engines

Custom-trained engines provide AI models fine-tuned with your specific data. This delivers key advantages over default generic engines:

  • Precise, contextual translations that accurately reflect industry and brand terminology
  • Consistent brand voice and style across languages
  • Ideal for translating industry-specific content containing unique jargon

By training models with your own data, these tailored engines ensure maximum relevance and accuracy for your content, maintaining your brand identity globally. List of supported languages for custom-trained engines

Custom engine training is an add-on service. Please contact success@helpshift.com to know more about custom-trained engines.

Enabling the Language AI for Agent Conversations

To begin harnessing the power of Language AI, you need to request the feature from the dashboard.

Follow the steps given below to request/enable Language AI for Agent Conversations:

  1. Log in to your Helpshift domain as an administrator.
  2. On the Helpshift toolbar, click Helpshift AI.
  3. Click REQUEST FEATURE under the Language AI for Agent Conversations option.
     
  4. You will see the confirmation dialog box on the screen.
Once the feature is requested, a representative from our team will reach out to you.

Upon enabling this feature for your domain, you will see the button change to FEATURE ENABLED.
 

Using Language AI for Agent Conversations

This section helps agents navigate the different functionalities of the Language AI feature.

Read end user’s message in the Agent’s language (English)

Agent can translate end user’s messages in any of the supported languages into English by clicking the Translate button next to the message.

After clicking on the Translate button, you can see the end user’s original message (in the original language) as well as the translated message in agent’s language (English).

For example, you can see the original end user’s message language is 'French', and agent used the Translate button to translate into agent’s language (English).

Agent can read messages from any supported language to English language.

Write in English and translate to end user's language

Agents can also write a message in English and then translate the same message to end user’s language by clicking the Translate button next to the message.

After clicking on the Translate button, you can see the message translated into end user’s language in the chat area.

After translating the message, agent can click on the REPLY button to send it to the end-user.

Now, agent can see both the messages in the chat. However, end user can only see messages in their language.

Note: Make sure that you click on the Translate button before sending every message to avoid sending the original message without translation.

Change the conversation language

This functionality is used when the end user is using a different language other than the language detected on their device. Now the agent has the option to detect the end user language and save it for translation.

To change the conversation language, follow the steps below:

  1. Select the issue where you want to change the language from what was detected.
    In the example below, issue language detected is ‘English’ and the end user's message language is in ‘Spanish’.

  2. In the agent conversation view, click the arrow (^) next to the Translate button.
     

  3. Under Language AI settings, click Select end user language.
    You’ll see the end-user language screen where "Default" is set to the issue’s detected language (e.g., English).
     
  4. Click the DETECT button.
    The system will automatically identify and display the end user's language.
  5. If the detected language differs from the default, you have to UPDATE the end user language.
    Your selected language will be saved and used for further translations until it is modified again for the conversation.

After changing the language, agent’s read and write message flow will be the same as mentioned above.
You can refer the section links as below:

Translated message can be same as original message if agent selects incorrect language from the dropdown. 

Troubleshooting for Language Detection

Language detection uses Language AI to identify the language of end user messages. In some situations, detection may return specific statuses or limitations. The following are the example cases to help you understand these scenarios.

ScenariosExample

Smart form – single message

Language detection requires more context. A single message from a smart form may not be enough.

Message in English (Default Language)

When the detected language matches the default issue language, no action is needed.

Message in a supported language 

When a Language AI supported language is used, it is accurately detected and ready for translation.

Message in an unsupported language 

Language is recognized but not supported by Language AI.

Message in a dialect of a supported language

Dialect messages are mapped to their closest supported parent language (e.g., Brazilian Portuguese).

Not enough messages to detect language

If the message volume is too low or lacks clarity, language detection fails.

Spanish written in English script

Phonetically written Spanish (e.g., "cooomo estaaas") in English script may confuse detection and lead to an unsupported status.

End user experience

End-users will have seamless experience as if they are interacting with an agent in their native language (For example, Spanish in the below image).

 

 Important to know

  1. Agent Language Support: Supported Agent language is English.

  2. Availability of Translation Feature: This feature is available for ongoing or open issues only. It is not available for already resolved or rejected issues. However, if translations were performed while the issue was open, you can still view both the original and translated messages for resolved or rejected issues also.

  3. Language AI uses metadata language as end-user’s language by default. Please make sure you enable language detection flag all the times in your issue language settings page to ensure metadata language is correct.