Intent Detection is a feature that allows your AI chatbot to understand the goal or purpose behind a user’s message. Instead of simply answering questions, the bot can take action — like assigning the conversation to a human agent or launching an input form — making it work like a real assistant integrated with your team and tools.

For example, if a user writes “I’m having trouble logging into my account”, the bot can detect this as a technical issue and either guide the user through troubleshooting or assign the chat to a technical support agent. You can use intent detection campaigns as complements to AI trainings created with URLS, Files and FAQs. Let’s use this same scenario to show you how they work.

Preparing the tools

For this tutorial, I created a Team Role for Technical Support Agents and assigned teammates to it. This will allow the AI to automatically assign relevant conversations to them.

I also made sure my Team Members have access to the account where we will place the AI Assistant with the Intent Detection Campaign.

I created a new flow as well and dragged an user input flow component. This allowed me to design a series of questions for customers wanting to request a feature. You can see the final result on the image below:

I made sure that the customer’s answers are stored in custom fields for easy later consultation. Check this article for more information on how to do it.

Finally, I created an AI Training Campaign. I uploaded all the information about my business that the AI Assistant needs to answer general questions.

With all this ready, we can proceed with the Intent Campaign.

Creating an Intent Campaign

Set Up the AI Assistant. Go to the Bot Manager on the social network you prefer; here I am using WhatsApp. Then select an account and go to AI Assistant option. Turn on the AI Assistant feature and select your AI training campaign.

Next, turn on the toggle button for Intent Detection. Click on Create to start a new Intent Detection campaign.*

Name the campaign. In this example, I used “Technical support”.

Then, write a few sample sentences that the user might use in the conversation with the chatbot for this intent. In this case, we can use these: “I’m facing an issue with your software”, “I need technical assistance” and “Can someone help me resolve a problem?”. Three to five examples, separated with commas, are enough for the AI to understand what the intent should be.

Now, select the dedicated toggle button for agent assignment. Choose your desired team from where the AI will randomly assign an agent. Write a response message to ensure the user that a human agent will be in contact soon, and click on the create button.

Let’s create a second intent detection campaign, this time for users that want to request a feature. Click on the Create button once again. Name the campaign and once again, provide some sample sentences for the AI to identify the intent. Let’s use “I found a bug”, “Can I request a new feature?” and “Are you planning to add new functions?”. This time, activate the toggle for launching a bot, and select the User Input Flow we created before. Save.

Finally, we will create an intent detection campaign for users requesting a refund. Give this campaign a name, write down some example sentences like “What is your refund policy?”, “How do I request for refund?” and “Can I get money back?”. This time, we will toggle the AI driven responses option, so the bot can answer by itself. We will create a new AI training campaign.

Click the Create Button to generate a new campaign. Keep the prompt message as it is.

Then, go to the URL option. We will train the AI with the policy refund on our website.

You need to copy the element selector of page and paste it on the training setup page. Check this article to learn more about training the AI with an URL.

Take the website’s link, paste it on the corresponding field as well. Generate a raw response and save.

Now, go back to the intent detection campaign page. Refresh the AI training data source and select the newly created campaign. Click on Create to save.

Finally, select the three Intent Detection campaigns we just created. If you do not see them, click on the Refresh button to reload the list.

Make sure to save your changes. Configure the rest of the AI Campaign. Toggle the “AI as Fallback only” option and optionally, enable the contextual memory and typing indicators. Click on the Save Settings button at the bottom.

Now, let’s test! Open your WhatsApp and type a message. In this example, we will write “I’m facing an issue with your software”. As expected, the bot understood and replied exactly as we pretended. You can see on the image the message indicating that the conversation was assigned to a human agent.

If you go to the Live Chat, you will see that this was indeed true. The conversation was assigned to a member of the Support Team we selected earlier.

Now, let’s ask for a new feature. As expected, the User input flow is launched and the customer is guided to provide details.

After answering the questions, they are stored on custom fields that the team can later check. You can access them on the Subscriber Manager.

Finally, let’s request a refund. Here, the AI was able to answer based on the site’s policy.

And that’s it! Play with intent campaigns to complement your AI trainings and turn your chatbots into real team players for the best customer experiences.