Set up your AI Generative bot

Generative AI bots, powered by our trusted generative AI, can create their own answers based on the information and instructions you provide. Here's how to create a Gen AI bot that will answer your customers' questions on its own, or assist your advisors.

 


Introduce it to your brand

First and foremost, your bot refers to your project parameters to find out on whose behalf it speaks. It's therefore crucial to check that your brand name is correctly entered in your project parameters, and, if necessary, to fill in the dedicated fields.

 

 

Give it a persona

In the bot builder, create a new bot. In the "Identity" section, set its name, appearance, description, language and transfer parameters.

 

Go to the "Scenario" section and select a "Generative AI" card.

 

Write its very first message, in which it introduces itself to visitors. We advise you to make it clear that he's a generative AI and that he's not infallible. Adapt the tone of this message to the one you're going to choose for the bot.

 

Select a tone: neutral, formal or friendly. Choose it wisely according to the habits of your industry (you'll have the opportunity to test your bot later).

 

 

Examples of tonality

Here's the same sentence in 3 tones:

neutral: "Yes, this dishwasher is Wi-Fi compatible and can be connected to the Internet."
formal: "Indeed, this dishwasher is Wi-Fi compatible and offers the possibility of an Internet connection."
cool: "Yes, this super dishwasher is totally Wi-Fi compatible, and guess what? It can be connected to the Internet! 😄"

 

 

Provide it with information (Knowledge)

To give you full control over your bot, you provide it with the information it needs to respond to your visitors, by selecting an AI knowledge that you've set up in advance. This knowledge will serve as a reference for your bot, enabling it to respond autonomously to visitors.

 

Restrict his behavior in "borderline cases

To prepare for any eventuality, you can set custom actions if the following events occur during the conversation:

  • visitors explicitly ask to speak to a human
  • the bot doesn't know the answer (because it's not in the resources available to it)
  • a prohibited word or expression appears in the conversation
  • a technical error occurs

 

Depending on the size of your teams and your overall strategy, you can use one of 5 solutions:

  • write a specific message: you have complete control over what the bot says, and write a message specifically for this situation, before the bot resumes the conversation on its own.
  • propose a transfer: write a message and plan a transfer to a distribution rule. The bot will only carry out the transfer if visitors accept its proposal.
  • transfer directly: write a message and transfer the conversation to a distribution rule, without asking visitors for their opinion.
  • continue to another step: write a message and continue the conversation with another card.
  • close the conversation: write a message informing visitors that the conversation is about to end.

 

Prohibited words and expressions

You can enter a list of prohibited words and expressions (= several words).

If a visitor's response contains an exact match to an item on this list, or if the bot is about to give a response containing one of these items, the personalized action is triggered, either in response to the visitor, or instead of the bot's spontaneous response.

 

For example, if your company (in our example, Supershopper) lists the name of your main competitor, if your visitor says "I think this product is cheaper at Wondershopper", the bot will immediately respond with the phrase you wrote for him: "I'm sorry, I don't know how to answer that question".

 

Test your Generative AI bot

Don't forget to test your bot's ability to respond to your scenario using the preview to the right of your scenario.

 

By clicking on the banner at the top of the preview to modify the test conditions, you can enter custom data corresponding to a product about which you wish to test the bot's knowledge, thus testing it in real-life conditions.

 

Here's how to enter custom data into your bot: