Articles on: Crisp Chatbot

How do I create a FAQ Chatbot powered by AI?

The Crisp chatbot provides an AI feature that is useful for sharing relevant FAQ articles automatically. An extremely useful feature that can assist your customers even when your support is offline. This article will show you how to create an FAQ chatbot that leverages AI to suggest the best help articles..

In this guide, we will first show you how to create a simple FAQ bot powered by artificial intelligence. This will help you understand how the New User Message block works in conjunction with the Search Help Desk block. It will also describe how the Search Helpdesk block functions and which helpdesk locales will be used. The second part of this article will guide you to create a more complex faq chatbot, by responding when no articles are found and humanizing the scenario.

To go to your Bot plugin homepage from the Crisp Workspace:
A. Go to Plugins
B. Search for "Bot" or click on "Bot" as it's the first plugin visible at the top left of your screen.

Create a simple FAQ chatbot.

1. Initiate the chatbot scenario

Firstly, create a new bot scenario and rename it. In this case, we will name it “FAQ Chatbot”. Add a new event block called New User Message and connect it to the Entry Gate.

The New User Message block will trigger on any new message, it will also fire when a file or audio message is sent to the conversation. If the user sends a text message, the message content will be memorized in the scenario, this is important for the next step.

2. Search FAQ articles with the users message

Next, add a new action block from Actions > AI Actions > Search Helpdesk and link it to the New User Message block. This is where the magic happens: the Search Helpdesk block will use the memorized message content from the previous step to search and find any relevant FAQ articles. If any help articles are found, they will be listed and sent to the user, otherwise, the block will not continue.

If you have multiple helpdesk locales, the block will first search for any relevant article based on the user’s locale. If no articles are found the configurable default helpdesk locale will be used, in this case, we will use the English locale as the default. Additionally, there is the option to edit the message prefix, in this case we will leave the default message.

Note that selecting the option 'All' will send any relevant faq articles no matter the locale.

Create a more complex Frequently Asked Question bot.

1. Only trigger bot on new conversations

In the first example, the bot template works but it could be improved. For example, you may only want your bot to search for FAQ articles if the conversation is new, this can easily be fixed with a condition block to check if the conversation is new.

2. Prevent from searching for FAQ articles on first message

You may have noticed a second issue. It's very likely that your user's first message will be a greeting, such as “Hello” and you might not want the bot to search your help desk for an article about a greeting. To fix these two issues, we will add three new blocks to our existing scenario;

A condition block to check if the conversation is new
A greeting message to ask the user to send their question
A second New User Message to memorize the user's question

Your workflow should now look like the image below:

3. Handle cases where no articles are found

If the artificial intelligence could not find any relevant articles, you may want to let the user know and direct them to a support agent. To do this, we will simply attach a message block to the second New User Message block. Make sure the branch priority is after the search helpdesk, to ensure the bot attempts to search for helpdesk before responding with the fallback message. More about branch priorities in this article.

4. Add some automation and a human touch to your AI Chatbot

Finally, let’s add some automation to resolve the conversation if the answer was helpful to keep the inbox free from clutter. Otherwise, we’ll leave the conversation open and let the user know an agent will be with them as soon as possible. To top it off, we’ll add show compose blocks to give the scenario a human touch.

That's it! You now have a AI bot to suggest FAQ articles to your users!

Updated on: 11/03/2024

Was this article helpful?

Share your feedback


Thank you!