How to configure conversational touchpoints

There are several aspects to be defined in the touchpoint UX that are unique for conversational touchpoints, on web, messengers such as Facebook or within the app.

Define the behavior in case an answer is not found

  1. In the Bold360 ai Console, go to Touchpoints > Widgets.
  2. Click Personalize under Floating widget.
  3. Select the knowledge base that you want to change.
  4. In the Error answers field on the Conversation tab, type the answer when an answer is not found. You can define either a text to be presented or an article ID. Your defined responses will be presented sequentially when the bot does not find an answer.

Best practices to consider:

  • On the first attempt, encourage the user to rephrase.
  • If you want to offer channeling only on later attempts, for instance on the second attempt, you can insert an article's ID, which is associated with a channeling rule.

How to define conversational UX

The definition of the web conversational touchpoints and the Messenger UX are quite similar. To define the Facebook Messenger user interface, do the following:

  1. In the Bold360 ai Console, go to Touchpoints > Widgets.
  2. Under Conversation configuration, click Facebook.
  3. Select the context information, the relevant knowledge base, and configure the touchpoint as follows:
    Option Description
    Facebook page This field is automatically updated once the Facebook page has been integrated.
    Persistent menu value To define the contents of the menu that is presented on the search bar, define header:article ID. Separate options with a comma. You can also define text that will be used to search once this option is selected. So for instance: Contact us: 1235677, Find a store: 1846649.
    Welcome article id Select the article that is displayed as a welcome message
    Enable Channels The bot will leverage the channeling rules.
    Enable Feedback To enable the bot to solicit feedback by prompting "Was I helpful?" Yes/No.
    Time before feedback Once feedback is enabled, the time passing until the user is prompted with feedback options. Generally set to 45-60 seconds.
    Ask an agent A persistent channeling option will be presented under each bot response.
    Show support center link A persistent link to the answer in support center will be presented under each Bot response.
    Enable answer in parts When the answer is longer then defined by Facebook's UX guidelines (e.g. 640 characters), the user will be offered a read more option.
      Capture a consistent context type and value to differentiate the content on the Bot from the content on other type of touchpoints. Syntax is context type: context value. For instance FB:yes.
    Search precision The search precision determines the NLP scoring threshold. The higher the precision the lower amount of answers will be presented. The Normal precision is the threshold that is used for regular search UX interfaces such as the support center search. The best practice for Bot is restrictive.
    Multiple header results The text that appears in case the user is asked to select between a few optional search results.
    Error answers Add a statement or article the bot shows in response to questions it cannot answer. The first answer in the list is shown in response to the first question the bot cannot answer. The second is shown in response to the second question it cannot answer. This is defined per knowledge base, not per touchpoint.

The definition of a web conversational touchpoint is quite similar. On top of the options listed above it has an option to restrict its usage by an API key. The configuration is performed on the Conversation tab in the floating or embedded widget configuration. For more information, see What are the options for configuring a conversational widget?.