BotCommerce’s AI Assistant is an advanced AI powered by OpenAI, designed to help you create more human-like interactions with your customers. Unlike traditional chatbots, it leverages NLP (Natural Language Processing) to understand user queries and respond with accurate, meaningful answers. This is not just a chatbot—it’s a comprehensive solution for customer service, marketing, and support automation.

The AI Assistant learns about your business and engages with your customers across WhatsApp, Facebook, Instagram, Telegram, and your website, providing faster  multi-language responses and smoother, more engaging interactions 24/7.

You can utilize this feature in two ways: through the Bot Manager’s AI Assistant or via the No-Match Template found in your chatbot’s action buttons. Each option offers slightly different functionalities for your AI Assistant. In this article, we’ll explore both approaches and guide you through the setup process.

⚠️ IMPORTANT: Before beginning this tutorial, you must input your OpenAI API key into the app, otherwise, the AI will not respond. Learn how to do it here. 

Bot Manager’s AI Assistant

An AI Assistant Training Campaign allows you to train your chatbot with tailored data for seamless, user-specific interactions. Key features include:

• Central AI Control Panel: Manage AI settings, enable/disable features, select training campaigns, and assign team members from a single hub.
• Smart Memory and Filtering: Use Contextual Memory, Restricted Topics, and customizable responses for efficient, focused performance.
• Custom FAQs: Add personalized FAQs directly to the Bot Manager for business-specific answers.
• Web Data Fetching: Train AI to extract relevant FAQ content from URLs, selecting sections via ID or Class selectors.
• File-Based Training: Upload PDFs, Word documents, or spreadsheets to generate FAQs automatically.
• Effortless Integration: Easily integrate your trained AI into chatbots for context-aware responses across all channels.

In short, this powerful tool ensures your chatbot provides accurate, dynamic responses tailored to your business needs, all without requiring any programming or manual prompting.

To access to it, login to your account. Once logged in, go to the AI assistant option of any of the platforms. For example, here we are using Facebook.

This is where you have to create a new AI training campaign and add all the data. Once your data is in there, you can test and tune the way the chatbot responds to users. Upon approval, you will be able to roll out your chatbot across multiple channels – on your website, WhatsApp, Facebook Messenger, and Telegram.

Now that you know the basics of getting started, let’s walk through the practical steps of creating your first AI Training Campaign.

Create a New Campaign. From your Bot Manager dashboard, click on the AI Assistant tab and enable it. Then, select the Create Campaign option. This will open a new  tab from where you can start the process of setting up your customized training data.

Set up a system prompt for the AI Training Campaign. Before adding training data, you’ll be prompted to name your campaign and provide a system prompt for the AI. We recommend giving the AI an identity (e.g., ‘You are our support agent’) and outlining specific steps for interactions, including response generation guidelines. Keep it concise; your business information will be added to complete the prompt later. When ready, click on Save.

Click the “+” button to add the training data. It will open a new tab where you can provide files, URL, and manual FAQs to train your AI. Let’s see how each one works.

Adding Manual FAQs

The basic way to add more content to the AI training campaign is manually inputting FAQs. Write the question first, put a line break and add the answer. You can provide as many FAQs as you want by clicking the add more FAQs button.

Don’t forget to press Save when you are done.

Adding an URL

The URL input method gathers training data by extracting content from a specified website using a CSS selector. This approach requires basic knowledge of website structure.

To begin, navigate to the URL you want to use for training data in your browser. Open the developer tools (inspect element), and identify the specific element to extract data from.

Enter the CSS selector for that element in the Selector Name box. Similarly, you can exclude sections of the website by specifying their selectors.

Revise the FAQs that will be generated automatically. Edit them if needed and save.
You can repeat this process to add as many URL elements as you need.

Adding Files

You can upload text, DOC, and PDF files (up to 2MB) to enhance your AI training.

Add a file. In the right panel, navigate to the Files section and click the (+) New button to upload a file from your computer. Once uploaded, click “Generate FAQ” to create FAQs.

Review and edit the FAQs as needed. Finally, save your changes. Repeat the process as many times as you need to.

Adding and configuring the trained AI campaign

Once you complete the training, you must add the your AI Assistant to your accounts, and configure its scope. Let’s see how.

Go back to the AI Assistant section. Select the AI Campaign you created before.

Specify the restricted topics. These are keywords that the AI should avoid responding to. You can add several, separated with commas. You should also add a message to be sent when the AI detects restricted keywords in the query.

Optionally, enable contextual memory. This will allow the AI to remember the user’s messages for as many conversations as you indicate.

You can also enable the AI to transfer the conversation to a live agent. When a user requests to speak with a human, the AI will deactivate, ensuring a smooth transition for seamless support.

Finally, you need to decide the scope of the AI. This is a crucial decision, since it will affect the whole dynamic of the chatbot. You have two possibilities:

• Set the AI Assistant for all queries: When this mode is selected, the AI will respond to all incoming user queries, regardless of whether they match predefined bot flows or templates. This is best for open exploratory interactions where users might ask a wide range of questions. Ensure your AI training is robust to handle a wide variety of issues; note that the No-Match template and any keyword-based flows will be ignored.
• Set the AI as a fallback only: When this mode is selected, the AI will respond only when the bot fails to to match any predefined flow or response template. It’s best for chatbots with well-defined keyword-based flows, where AI is used as a safety net for unexpected or out-of-scope queries. Use this mode to minimize AI interactions and maintain a structured chatbot experience. Note that if the No-Match template is turned on, it will have preference over the AI Assistant responses, so we recommend to leave it off when you are using this option.

Once you have finished configuring the AI Assistant, make sure to save the settings to apply them. Then, you can add this same AI Campaign throughout different accounts in different channels. All you have to do is go to the AI Assistant section of said account, enable the AI Assistant, select the same AI Training campaign, and set up the scope, restricted keywords and memory settings as desired.

No-Match Template training

The AI Assistant training is ideal for simplifying the process by letting the AI answer customer questions about your business without complex prompting. However, there are times when you may need more flexibility in crafting responses. For example, consider a scenario where the AI is trained to select and provide a tutorial that addresses a customer’s query:

In such cases, using the No-Match template for training might be more beneficial, as it allows you to input a single, structured text prompt tailored to your needs.

To set it up, you will first go to Settings > OpenAI (the same place where you placed your OpenAI API Keys). On the Instruction to the AI field, provide the general system prompt. Note that this will be valid for all the accounts that are using the AI with the No-Match template method. Accounts with an AI Assistant Campaign will skip it and use its own system prompt instead.

Note that, once again, the system prompt is rather brief, containing just the interaction guidelines. The rest of the information will be added on the No-Match template.

Go to the Action buttons section inside the Bot Manager of the account where you wish to install the AI. Double click on No-Match. As context, in a conventional chatbot, the No-Match template is a flow that gets triggered when the user’s message does not contain any of the keywords recognized by the bot.  By placing the AI training on the No-Match template, we will allow the AI to answer any query not associated with another flow. As we mentioned before, if the No-Match template is enabled for an account, it will take precedence over the AI Assistant Campaigns discussed on previous sections.

The No-Match template flow will open. As you can see, right now it triggers a Text message. Delete that Text Component.
Now, drag an AI component into the dock and connect it to the Start Bot component. Double click on it to configure it.

Paste the prompt containing your business information. Note that is is structured as a sample conversation in text.
You’ll also notice that there’s a Open AI training box below the space for the training. This would allow you to link the No-Match template to a training created with the AI Assistant instead. Since that would replace our prompt, we will leave it empty.

Apply the changes. When ready, save the component, and then save the flow.

Finally, we need to turn on the No-Match template. Go to Bot Manager  > Settings. Turn on the No Match Reply switch and enable it for Every Time.

Done! Your AI must now be working for this account. If you want to share the same training on other channels or accounts, you just need to paste the training prompt on their No-Match Template as we did here.

Mixing both methods

Before we said that if you have a No Match Reply On, it will take precedence over the AI Assistant when the last one is set in Fallback Only mode. However, there’s a way in which you can take advantage of this, by placing that same AI Assistant in the No Match Template instead of a text prompt.

Basically, you would have to repeat all the steps we saw before to to set up the AI Assistant. That is, create a new campaign; train the AI with files, URLs and FAQs; enable the AI Assistant for the account, and configure its scope. Then, you simply turn on the No Match Template with whatever frequency your prefer.

Create an AI Flow. Go to the Action buttons section, click on the No Match option to configure it, and as we did before, replace the Text component on the flow with an AI Component. Double click on it to configure it. Just this time, instead of pasting a prompt, leave it empty and select your AI Training campaign on the corresponding box.

This way, the chatbot will go through your keyword-based items first, and all the rest of the messages will be answered by the AI.

Play with all the elements to build dynamic chatbots tailored to your needs. And if you require a little help, our experts can help you design the perfect chatbot here.