In this article, we show you a practical case to learn how to use Artificial Intelligence within your conditionals.
📍 Practical objective: Detect whether the user intends to consult or purchase a spare part (regardless of what exact words they use), to automatically direct them to the corresponding branch of the flow.
⚙️ Step by step: How to configure Atom's Magic
1️⃣ Insert the Evaluate response component From the left sidebar menu in Flowbuilder, select the Evaluate response component and drag it to the canvas, placing it right after the message where the customer will write their query.
2️⃣ Activate Atom's Magic In the right configuration panel of the component, activate the Atom's Magic button (identified with the magic wand icon ✨). This disables rigid evaluation by exact keywords and enables evaluation by intention understanding.
3️⃣ Configure the main branch Define the basic parameters so the flow knows where to direct this intention:
Branch name: Spare part
Value to compare: Message body
Condition: is equal to
Expected text: Spare part
4️⃣ Train user intentions (The Prompt) In the Evaluator training section, you must enter the expected scenario so that AI understands the context of the "Spare part" intention.
Suggested description: "Enter here when the user's input or message has a clear intention to buy, quote, search for, or ask about a vehicle spare part or a spare part in general."
Examples to train the AI: You can provide common phrases.
🔁 Tip: You can use variables like
{spare part}and{model}in your examples so the AI learns to recognize flexible patterns (E.g. "I'm looking for the price of a {spare part} for my {model}").
5️⃣ Connect the output to the desired action Connect the new Spare part branch to a Message, Assignment component, or any action you wish to perform. You can use this instance to offer a category menu, provide an automatic quote, or transfer to a spare parts specialist.
💡 Additional tips for better performance
➕ Add more intentions: You can add more branches within the same evaluator with other intentions (such as "Workshop appointment" or "Reschedule"), following this same configuration pattern.
🧪 Run tests: Use the simulator with different phrases, synonyms, and spelling errors to ensure the AI is correctly interpreting the intention.
⚠️ Avoid confusion: Don't train many intentions that are too similar or ambiguous within the same evaluator if they are not clearly differentiated in their descriptions, as the AI could get confused when choosing the route.
Your bots now understand context! 🚀
By using Atom's Magic, you overcome the limitation of exact words and offer a much more natural experience. Train your branches intelligently and take your automation to the next level. ✅