Let’s scope it out together!
Using intent management dashboard, we introduce the intents to the chatbot, add some sample user utterances, attach some context, and tell the chatbot how to respond to the intent.
After the training portion, we can test the intent and see if we hit an adequate confidence score (> 80%) for that intent. Upon failing, we need to add more utterances — or, give it more data — until the chatbot can confidently distinguish that particular intent from other intents.
Two cases where human attention is needed the most: when the conversation goes triggers the fallback mechanism, and when the bot detects a negative user sentiment.
In that case, you can take over your chatbot’s conversation, type in your answers and continue conversing with the customer on the bot’s behalf. While you handle the conversation, the chatbot uses machine learning to learn how you respond to that particular intent.
There is so much information we can take from a conversation. It’s just a matter of how we process it and turn it into insights. Other than that, your goals largely affect the metrics that you should be looking at — the way you interpret the metrics might change, too!
Here at Botbot.AI’s data analytics dashboard, we present you the usage level of your chatbot, popular topics among your customers, as well as how well your chatbot performs. Coupled with the right analytics, you can then take concrete steps to improve or fine-tune your business processes.
Experience the added productivity you can get with chatbots