Easily set up a chatbot with Watermelon
Watermelon makes it easy to set up and start using your own chatbot in just five steps: adding knowledge, training, testing, publishing and optimising. Read on to see just how easy each of the five steps actually is!
The strength of a chatbot lies in the depth of its knowledge base. In order to understand the intent of the questions a customer is asking, and to provide the correct answer, your chatbot will need a basic amount of knowledge. But how can you make sure your chatbot understands what’s being asked?
When you’re setting up a chatbot with Watermelon, you don’t need to know any programming languages. All the knowledge that you’re adding is textual, natural language. The entry fields make it clear what you should enter and where.
Your chatbot must learn at least three variations of the same question. Your customers ask the same question in different ways, and naturally you want your chatbot to understand all of them. The chatbot learns the variations of the questions and creates a database with its experience in order to learn the intent of the question. Even if the formulation of the question is not a 100% match with one of the variations you’ve entered, the chatbot can still provide the correct answer.
You can also add knowledge to your chatbot by using the Conversations module. Build your own conversations using our easy-to-use answer trees, also known as conversational design. Ask customers follow-up questions based on the answers they provide.
By using conversations, you can do much more than just have your chatbot provide information. Using conversations with your chatbot can generate leads, plan appointments, send PDF files and much more!
Each time you’ve added knowledge to the chatbot, you’ll need to train it. During the training session the chatbot analyses the knowledge you’ve added. This links the questions and all of their possible variations to the correct answers.
With one click of a button, the training starts in the background. You can do other things such as look at your conversation statistics, add new channels or edit user permissions – all while the chatbot is training. The chatbot will indicate when it’s ready for the next step.
Now that your chatbot has knowledge and has been trained, it’s time for testing! Testing happens in a closed chat environment where you have a one-on-one dialogue with your chatbot. You don’t need to publish your chatbot online in order to test it. Ask it the new questions you’ve taught it in as many ways as you can. If you don’t like the answers the chatbot is giving you, go back to step 1 and see what you could improve. If you’re satisfied, it’s time to go live with your chatbot!
If you’ve successfully completed the first three steps, it’s time to publish your chatbot live online. This is the most exciting part, because your customers are going to get to meet your new digital employee. Once the chatbot is live it starts working immediately, helping an infinite number of customers simultaneously.
You can have your chatbot live and functioning on multiple channels at the same time. You won’t need multiple chatbots in order to use a chatbot across channels.
Your Watermelon chatbot can be live on:
CXhub allows your customers to converse with your chatbot on your website. Offer chat on your website to reduce the number of emails and phone calls your customer service team needs to field. Customers will be even more engaged with your company and brand once they see how easy it is to have a conversation on your website (that’s powered by CXhub from Watermelon!)
Start chatting with your target audience on the channel many of them are using to chat with each other. Start using your chatbot on Facebook Messenger to easily expand the reach of your customer service team to channels which might be new to you, but not new to your customers.
Your chatbot can respond to incoming Telegram Messenger conversations. Being available on this channel makes your customer service even more accessible.
You can use your Watermelon chatbot to respond to private messages on Twitter, which is especially handy if this is typically a channel that generates a lot of traffic.
Once your chatbot is live and is having conversations with customers, the final step begins. In the first step you gave your chatbot some basic knowledge. However, your customers will likely surprise you by asking questions you didn’t think of in ways you didn’t think of. Of course your chatbot is up to the task of learning how to answer these questions too. The longer it’s live, the more intelligent it can become.
The Watermelon chatbot records the questions it could not understand, and presents these in a ‘Mismatches’ module. You can resolve the mismatches easily and add this to the existing knowledge base. The input of customers is key to optimising your chatbot over time.
After you’ve added the knowledge from resolving mismatches, you can train and test the chatbot again. This time you’ll know for sure those questions which were once new are now old hat!
Optimisation is a continuous process that takes more time when your chatbot is the new bot on the block. The longer your chatbot is live, the fewer mismatches you’ll need to resolve. You’ve got the power of Watermelon at your fingertips to easily create the optimal chatbot!