This video walks you through the complete process of setting up a generative AI chatbot from scratch inside the Kommunicate dashboard. Starting with bot creation and configuration, the tutorial then dives into Kommunicate's generative AI specific features showing how they work in practice, directly inside the platform.
Whether you're evaluating Kommunicate or just getting started after signing up, this is a practical, visual guide to understanding how the GenAI capabilities come together from initial setup to a fully configured bot ready for deployment.
Welcome to Kommunicate. Today we'll be exploring how Kommunicate can revolutionize your customer service experience. To get started, you'll need an account. If you don't have one, you can sign up using the link provided in the description below. Once you're logged in, navigate to the Bot Integration tab to create a new bot.
You'll find the Create New Bot button there. For the purpose of this tutorial, we have already created a bot named Kompose AI Bot. Let's go to this bot and click on the Bot Builder button. This will take us to the Kompose Bot Builder page where you can configure your bot.
On the left side you'll find several segments: Welcome Message, Intent, Small Talks, Knowledge Source, and Default Fallback.
We'll start with the Welcome Message. This message is displayed to the user as soon as they interact with the chatbot. Here you can configure your welcome message. To see how it will appear to the user, click on the Show Preview button. We've already created a welcome message. To add interactive buttons to your message, click on the plus button. We've already added a button named Card Carousels.
Next, we'll move on to the Intent segment. This is where you create responses to user queries. For instance, we've created a button named Card Carousel. To define this intent, go to Bot Says and specify what the intent should do. To create a custom payload, simply click on the More button, then select the Custom Payload option. We've added a rich message which will look like this. It's a Card Carousel.
We have a resourceful website, docs.kommunicate.io, offering a variety of rich messages for you to utilize. To use it, just copy the code and paste it into the Custom Payload section, then adjust it according to your requirements. When you've finished configuring, click on Train Bot to save your modifications and train the chatbot.
Now let's focus on the Knowledge Source. Here you can upload documents that you want the chatbot to learn from.
Moving on to URLs, you can also input the URL of any website to train your chatbot. For example, if you have an FAQ page on your website, just provide that URL and our chatbot will start training on it. You can even connect your Zendesk Knowledge Base or Salesforce Knowledge Base using this feature.
I have already uploaded a Kommunicate support PDF and the chatbot has been trained on it. Similarly, in the URL section, as soon as you enter the URL and click the Add button, the chatbot will start training on it.
This feature is highly advantageous as it enables the chatbot to learn from as many URLs as needed, a task that would be impossible for a customer service agent. Furthermore, there's no need to translate and upload documents or URLs in different languages, as our chatbot is intelligent enough to respond to questions in the language they were asked.
Finally, let's proceed to the Default Fallback. This is the message you want to display when the chatbot cannot answer a question. You can display a message like "We'll connect you to a human agent." Click on the More button and you have the option to hand over the conversation to a specific individual, a team, or another bot. Once you've set this up, click on the Save button.
Proceed to the Bot Settings and click on the Settings button. Here you'll discover Custom Instructions and a toggle switch.
Custom Instructions allow you to control the tone and limitations of your chatbot. You can choose a casual, friendly, or professional tone and decide whether the bot's responses should be short or lengthy. For instance, we've set a limitation that the bot should not respond to recipe related queries. You can turn this limitation on or off and then save your settings.
Contextual Understanding is another feature that enables the bot to better comprehend what the customer is asking and answer accordingly.
Next, let's look at the ChatGPT integration. Even without a specific intent, the chatbot will search the web for answers and provide them to you.
Now let's test our chatbot. Click on the Test button. You'll see the welcome message and the button we've configured. Clicking on the button will display the intent we configured in the custom payload using the rich message.
For example, even though we haven't specified anything about making a pizza, if we ask, the bot will search the web for the answer because we have enabled the ChatGPT toggle. The bot has provided instructions on how to make a pizza.
Now if we ask a question like "How long should it be in the oven?" without mentioning the word pizza, the chatbot will still understand that the question is related to pizza and provide an answer. This is the power of Contextual Understanding. Even though the word pizza wasn't explicitly mentioned, our system was able to recognize the context.