There are a number of non-configurable bot behaviors that are part of your bot’s functionality. These are triggered by the customer entering a free-text message into the widget or mobile SDK. The bot then assesses the content of the text entry and responds with one of the following behaviors:
These responses remain active in customized bots.
These automated messages appear in your selected bot language. See Managing languages for a bot.
Conversational shortcuts
A conversational shortcut uses AI to assess any free-text message entered into the conversation by a customer and automatically jumps to the relevant part of an answer.
Article suggestions
Article suggestions are only available if you have a publicly available help center. If no help center is available, the article suggestion behavior will not be triggered, and a fallback flow is triggered instead.
Fallback flow
The fallback flow is triggered when none of the abovementioned behaviors help the customer self-solve their support issue. The bot gives them the option to rephrase their question, or offers up to six possible answers for them to choose from.
You can configure the fallback flow. See Building a bot using answers: Fallback response.
37 Comments
I can see that one of our colleagues is already assisting you with regard to this concern. Rest assured that we'll do our best to help you with your inquiry.
Zendesk really needs the article recommendations to ask for feedback after recommending the articles. Having it automatically ask the "Was this helpful?" step and give the option for "Yes, my question is answered" vs "No, I still need help" would be both more helpful for our customers and also helpful for reporting purposes. This should be a standard feature and it should be included in the data in the Explore report for article recommendations.
The main issue here is that customers do not understand or know that they can still ask for help after getting the article recommendations since they're not being prompted. They just think the bot is "hanging up" on them and then they think they can't get any help, which is a pretty serious issue for something that is supposed to be more helpful to customers. Because of this, we've had to turn off Article Recommendations until there's a fix on this.
Please add a new step option for "Search Guide Articles" which triggers a full search against Guide.
The bot often gets confused if it can't find an Answer with a matching search term, and when "Recommend help center articles" is checked, it will often show articles which are completely irrelevant and off topic.
This would allow us to add the step for "ask if question resolved" after Help Center article searches,
Jason Kral have you had a chance to see these tips?
The above screenshot is at 1:15 into the video. Following these tips may help improve the Recommended articles. Also, you may want to consider adding labels to the articles.
About the Answers, how many do you have set up in your instance? How many training phrases have you set up for each Answer?
I agree that having another step option for Searching Guide Articles would be a great addition.
As per KROB's reply on July 7, 2023, when using the "Recommend help center articles" option, sure, it will present some articles to the user if it matches any, but then nothing else - there should be an option to ask the user if the articles helped, or provide a button to another "answer" so it can then step in to an answer bot flow.
So basically, how we present the user with the next step after it has recommended the matched help center articles?
Whats the ETA of asking if an article recommendation was helpful? Such a poor customer experience the way it is set up now. This is low hanging fruit the Zendesk team should easily recognize and fix.
When will restricted (segmented) articles be available in the search function? They are available within "Suggesting restricted help center articles in a bot answer".
This delivers a very different experience for a requester within a flow in a widget vs. the search of the widget.
Please sign in to leave a comment.