System Replies, unlike intent replies, are not used to answer any queries. Instead, depending on the reply, the purpose is to:
- Welcome customers - Welcome Reply
- Inform customers when there are no agents to take the chat - Failed Escalation Reply
- Inform customers that their query is not understood - Default Reply
- When the user writes in a language the virtual agent doesn't support, inform the user of the languages that are available - Unsupported Language
- Inform customers when there is some form of a technical issue - Technical Error Reply
- Collect feedback from your customers to measure how your bot performs - Collect Bot Satisfaction (BSAT) Rating
System replies are super powerful and should be included in the conversation design process to provide the best experience for the visitor.
By default, we suggest text, however, you should go in and adjust it to your Virtual Agents tone of voice, include some FAQ's or define alternate escalation strategies.
Welcome Reply
A Welcome Reply is a great way to set the tone when your customers reach out via chat. This can be set up in your CRM platform or Ultimate's Dashboard. The benefit of setting it up in the Dashboard is that you can build a holistic dialogue flow that is more intelligent than a standard text reply.
However, not all CRM platforms support this feature.
See the table below for details:
No action needed | "Trigger Reply" action needed | In CRM platform only |
|
|
|
Learn more about how to set it up here.
*Sunshine Conversation requires adding an additional snippet code to the widget. Learn more about it here.
Failed Escalation Reply
Failed escalation reply is used when no agents are available to answer the chat - this could mean that no agents are online, or the agents online were all busy and not able to take the chat.
Default Reply
Default reply is used when the bot is not confident to decide which intent should be triggered. The confidence threshold can be viewed in Bot Settings by Client Admins.
This is arguably one of the most important replies to consider from a conversation design point of view so that the user gets to a solution as quickly as possible and is not frustrated from being misunderstood.
Unsupported Language
Thanks to the AI fueling our language detection we can identify 109 languages based on the text a user sends. Based on the languages you have active in your bot, this will depend on which response is sent. When a user writes in a language you do not have an active reply in, this is the reply that will be sent.
Our recommendation, adapt the reply to include the languages the bot can communicate in so the user knows what the bot can handle, and what they can say if they want to be passed over to an agent.
Keep in mind that the unsupported language reply is sent when the language detection confidently identifies the language in the user's message to be any language that is not active in the bot. When language detection cannot identify a language with confidence, the bot falls back to the default language and the unsupported language reply is not sent.
Technical Error Reply
This is a reply that is triggered when there is some kind of error, either from the CRM or the Ultimate side. An example of when this can occur is when an unexpected error occurs when connecting to an external API.
You will be given a default message however, you can choose to adjust this to give further information or alternative communication options.
Collect Bot Satisfaction (BSAT) Rating
At the end of each chat conversation, you can ask your customers to rate their customer support experience on a scale of 1 to 5. Similarly to CSAT scores for human agents, BSAT is used to measure customer satisfaction with the service provided by chatbots. You can configure which resolution states can trigger the feedback prompt and how much time should pass before it appears.
Additionally, you can use BSAT scores to filter the Conversation Logs and look for areas to improve. On the Analytics page, you can learn more about your chatbot’s BSAT score and segment your data by language or intent.