Multilingual support

Understand multilingual support in Ada

Communicating with your customers in their preferred language makes them feel like VIPs. Ada lets you do just that! Use your AI Agent’s multilingual functionality to provide that VIP experience to each of your customers.

English is the default support language in all Ada AI Agents. You can’t disable it - your AI Agent will always be able to reply to English-speaking customers. Your AI Agent can also always understand customer questions in any of Ada’s supported languages, regardless of whether they’re enabled in your AI Agent, but should only respond in the languages that are enabled in the multilingual settings.

Inquiry translation and knowledge ingestion

For Ada to understand and respond to your customer inquiries in non-English languages, it will need to translate customer inquiries as well as your knowledge base content. Both of these translation types are carried out either using Google Translate, or via Native LLM translation.

For languages supported by Google Translate, customer inquiries will be translated into English before Ada’s Reasoning Engine generates a response that will then be translated back into the source language again by Google Translate before Ada responds.

For languages supported by Native LLM translation, customer inquiries in non-English languages are assessed and responses are generated directly in the same language by a non-English LLM model. This results in improved understanding and accuracy in the generated non-English response.

• For web chat and email, supported languages will see better response quality, with the AI Agent better reflecting nuances in your knowledge articles for that translation. Whenever possible, we recommend only using articles that are fully supported for knowledge ingestion with your AI Agent.

• Voice functions differently. While knowledge ingestion works the same way across all channels, voice conversations can only be enabled for explicitly supported languages. For more details, see Enable Voice in multiple languages.

Supported languages

Ada’s multilingual functionality supports the following languages:

For supported right-to-left languages, web chat is tailored for the best customer experience. This includes flipped UI components such as settings, loading bar, etc.

You’ll be able to create Ada articles in all the languages supported for knowledge ingestion.

Understand how multiple language support works

With multilingual support enabled, your AI Agent does the following:

1. Determines the language the customer is writing in and identifies how best to respond to the customer using Ada’s Reasoning Engine

2. If your knowledge base contains ingested content in the same language as your customer inquiry, the generated response will only leverage this matched language content when responding.

3. If your knowledge base does not contain content in the same language as your customer inquiry, the customer inquiry will be translated into English before the Reasoning Engine generates a response and translates the generation back into the customer’s original language.

4. If the language is not an enabled language, the AI Agent will reply in English.

Turn on multiple language support in your AI Agent

To enable multilingual functionality in your AI Agent, follow these steps:

1. On the Ada dashboard, go to AI Agent profile > Languages.

2. Under Available Languages, select any languages you want your AI Agent to be able to send messages in, in addition to English, which is always enabled.

Enable only the languages you need

Tempted to enable every language you possibly can? Hold off! Consider the following best practices:

• You should only enable languages that your agents speak. Otherwise, you might hand off customers to agents who can’t help them.

• For each language you enable, you should have someone on your team who speaks that language. That way, they can review conversations in the Conversations view to help gauge the AI Agent’s performance.

Set your AI Agent’s starting language

The starting language is the language your AI Agent starts web chats in. If most of your customers speak the same non-English language, you can choose to set it as your AI Agent’s starting language. Your AI Agent will automatically begin every conversation in this starting language, but of course, customers can still change the language at any time. The starting language for all new Ada AI Agents is English.

Add the language script to Embed2

You can change the bot’s starting language by making a small addition to the Ada Embed2 script. You might already be familiar with it, but if you need a refresher, check out our Embed2 documentation. Let’s take a look at how to set up an AI Agent’s starting language to French.

1. Find the default embed script on your webpage. It should look like this:

   <!-- <head> -->
   <script
     id="__ada"
     data-handle="yourbothandle"
     src="https://static.ada.support/embed2.js">
   </script>

2. To set the AI Agent language, add the following window.adaSettings snippet between the head and the script of the basic Ada Embed2 script, and add the language format code from the ISO 639-1 language format.

   <script>
     window.adaSettings = {
       language: "fr",
     }
   </script>

   Your revised Ada Embed2 script should look similar to this:

   <!-- <head> -->
   <script>
     window.adaSettings = {
       language: "fr",
     }
   </script>
   <script
     id="__ada"
     data-handle="yourbothandle"
     src="https://static.ada.support/embed2.js">
   </script>

Understand how your AI Agent initially determines the customer’s language in Ada’s web chat

Your AI Agent adheres to an order of operations when determining the chat language. It proceeds sequentially through the following criteria to determine the opening language:

1. Embed2 - If the Embed2 script defines a default language, the AI Agent will always launch in that language. If the language is not set in the Embed2 script, it moves on to check the customer’s browser language.

2. Browser - The AI Agent launches in the customer’s browser language. If the AI Agent cannot detect the customer’s browser language, it launches in English.

3. English - If both a default language isn’t defined in the Embed2 script and a browser language can’t be detected, the AI Agent launches in English.

Understand how your AI Agent initially determines the customer’s language in other channels

Other channels don’t have as much information available for your AI Agent to determine the customer language. In these cases, the AI Agent uses Ada’s language detection feature to determine the language the customer’s input is in.

Understand how customers can change the chat language

Customers can change the chat language using one of the following methods:

• Autodetect - If a customer types a message of at least three words and 20 characters in an enabled language, the AI Agent will try to identify the language and change the language of the chat.

  The three word minimum doesn’t include common words, such as “and,” “the,” etc.

• Language name - If a customer types the exact name of an enabled language (e.g., “French” or “Français”) into the chat field, the AI Agent will change the chat language accordingly. This must be an exact match — typos or the name of the language within a longer message (e.g., “Do you speak French?”) will not trigger a language change.

• Chat settings - The customer can click the Settings icon access the chat window settings in Ada’s web chat. They can then click Change language and select one of the languages you have enabled in your AI Agent.

Why didn’t my AI Agent switch languages?

There are several specific reasons why your AI Agent might not switch languages as expected:

• Minimum character and word requirements for Latin-based languages

  • For messages written entirely in Latin characters, the message must contain at least 3 words (space-separated) and 20 characters to trigger a language switch
  • This requirement applies to all languages except: Greek, Hindi, Japanese, Khmer, Korean, Myanmar, Punjabi, Tamil, Thai, Simplified Chinese, and Traditional Chinese

• Switch-disabled language pairs

  • Some languages are considered too similar by our model, so switching between them is disabled to prevent premature language changes
  • These pairs include:
    • Malay and Indonesian
    • Simplified Chinese and Traditional Chinese
    • Spanish and Catalan
    • Croatian and Serbian

• Switch-disabled languages

  • Once set to these languages, the AI Agent cannot switch out of them through language detection:
    • Haitian Creole
    • Tagalog
    • Bosnian

• Non-Latin writing systems

  • For languages with non-Latin writing systems (like Hindi, Chinese, etc.), we do not support recognizing Latin-based versions of these languages
  • For example, Hindi written in Latin characters (like “kya hal-chal hai”) cannot be detected, while the same phrase in Devanagari script (क्या हाल-चाल है) can be recognized

• Diacritics requirement

  • For languages that use diacritics (like Vietnamese), the model requires proper diacritic usage
  • Messages without diacritics may be misidentified (e.g., “voi” without diacritics might be detected as Italian instead of Vietnamese)

• Low performance languages

  • The model has lower accuracy for these languages:
    • Afrikaans
    • Catalan
    • Croatian
    • Haitian Creole
    • Malay
    • Norwegian
    • Slovenian
    • Urdu

• Unicode-based detection

  • For certain languages with unique writing systems, detection is based on Unicode rather than the model
  • This applies when the entire message uses only one of these writing systems:
    • Greek
    • Devanagari
    • Hiragana
    • Katakana
    • Khmer
    • Hangul
    • Myanmar
    • Gurmukhi
    • Tamil
    • Thai

• Denied words

  • Messages containing certain common words across multiple languages will not trigger a switch
  • These words include: “no”, “meta”, “variable”, “swish”, “support”, “agent”, “face”, “id”, “selfie”, “bet”, “confirmation”, “start”, “vacation”, “hold”, “nya”, “fuel”, “point”, “domain”

Work with custom translations

For the majority of conversations that take place in Ada, your AI Agent automatically translates all content. The exception occurs when you want to customize the translations in any structured piece of content in your AI Agent composed of blocks, like your greeting.

Within any piece of structured dialog, like a greeting, there are two translation methods available to you:

• Automatic translation – When customers ask questions in different languages, your AI Agent automatically sends your English content to Google Translate, and then responds with the translated content.

• Custom translation – Manually add your own translations. Adding a custom translation to structured content overrides automatic translation for that dialog in that language.

View and modify translated content for a piece of structured content

For pieces of structured content that don’t have custom translations, you can get a preview of how Google will translate the content. Then, if you want to modify the translation, you can do so.

1. Ensure you have enabled all the languages that you want to serve translated content for in your AI Agent. For more information, see Turn on multiple language support in your AI Agent.

2. On the Ada dashboard, open the piece of structured content you want to edit translations for. For example, to view your greeting, go to AI Agent profile > Greeting.

3. From the list of your AI Agent’s supported languages, select the language you want to see content for.

   The non-English languages can have two different headers above them, indicating the types of translations you’ll see:

   • Automatic Translations - Translated content comes directly from Google Translate and hasn’t been modified at all.

   • Modified Translations - Translated content contains custom translations and/or modifications to Google’s default translations.

   

   If you see that the language you want to use isn’t in the list, click Manage Languages to enable that language in your AI Agent.

4. If the dialog only has automatic translations, click Modify Translation. The blocks become editable.

   

   If the dialog already has custom translations, the blocks are editable as soon as you switch to that language.

5. Modify the translation as required and click Save. If the dialog didn’t have any custom translations before, the language now appears under Modified Translations in the list of supported languages for this Answer.

Remove a custom translation from a piece of dialog

To delete a custom translation from a piece of dialog:

1. On the Ada dashboard, open the piece of structured content you want to edit translations for. For example, to view your greeting, go to AI Agent profile > Greeting.

2. From the list of your AI Agent’s supported languages, select the language you want to delete the custom translation for.

3. Click Delete modifications. Then, in the confirmation window that opens, click Delete.

   

   Once you delete a custom translation, your AI Agent will go back to serving automatically translated content to customers.