Watermelon guide

Get started with Watermelon

Learn all about building chatbots and become an automation expert.

Table of contents

Building your chatbot

Great job! Your Watermelon platform has been set up and you can start with your very own chatbot! Before you take the plunge and start building your chatbot, it’s a good idea to do some preparation first. Let’s begin with creating a scope.

making conversations with watermelon view

Creating the scope

The very first thing to do when building a chatbot is to decide what you want your chatbot to do. Think about which questions you want to answer, which information you want to give your customers, and how you want your customers to navigate through conversations. This overview of what your chatbot should be able to do, is called the scope.

When you’re preparing your chatbot scope, it’s good practice to divide your scope into multiple versions. Version one could be a more basic version of your chatbot that answers the most frequently asked questions. After that you can move on to version two, where you can incorporate more specific questions and topics into your chatbot. Eventually you can go even further and move on to version three and maybe even version four where you can integrate your chatbot into other systems or add more channels.

It can be easy to dive in too quickly and lose track of what you’re doing and what you want to achieve. By using multiple versions of your chatbot scope, you keep the implementation manageable and easy to oversee. Another benefit of working with multiple, smaller versions of your chatbot is that you can easily gather feedback in between versions. This way you can quickly and simply improve your chatbot as you go!

Conversational design

With the scope determined, you can finally start building your chatbot! So let’s talk about how you can construct your chatbot conversations. Let’s take a look at Conversational Design.

Now that you know which questions you want your chatbot to answer, it’s important to consider how customers might ask these questions and how they go about looking for the answers. When you’re building your chatbot, try approaching it from your customers’ perspective. Think about the way they navigate through your website, which topics they encounter and which questions might pop up along the way. With your customers’ journey in mind, it becomes a lot easier to construct a conversation around that.

Instead of immediately building conversations, it’s useful to sketch out the conversation flows first, either on paper or using digital flowchart tools. Sketching out your conversations first can greatly help you get a good overview of your customers’ journey through the conversations. This way you can easily see which decisions customers can make during a conversation and which questions they can get answered.

After you’ve mapped out your conversations and you have a clear overview of what your conversation will look like, you can start building your chatbot in Watermelon.

Chatbot setup

To get started with your new chatbot, log in to the Watermelon platform and click the Chatbot icon in the top menu bar. If it’s the first time you’re opening this menu, you now get to see a welcome screen with three steps for creating your chatbot. The first step is to simply click the Start button.

The next step is where you can form the personality of the chatbot. The personality is a crucial part of your customers’ experience with your digital assistant. This covers everything from the chatbot’s name, its face, the tone of voice, and even to smaller details such as personal interests and hobbies of your chatbot persona. It’s important to carefully consider the personality of your chatbot, so that it matches the rest of your companies’ brand identity and tone of voice. After all, your chatbot will be the first impression your customers will get of your service, so you want it to be a good one!

On the second screen of the chatbot setup you can enter a name for your chatbot and upload an avatar. In the grey text fields you can decide how your chatbot will respond to certain situations. For example, what will your chatbot do when someone asks a question it doesn’t know the answer to yet? Will it ask the customers to rephrase their question, or will the chatbot forward the question to a live agent? Or what if someone asks the chatbot an unrelated question, such as “What is your favourite food”? Will your chatbot ignore the question, or will it play along and give an actual answer? The way your chatbot handles these situations is integral to the customer experience.

Once you’ve set up the personality of your chatbot, you can scroll to the bottom of the page where you can see the grey Continue button. All you need to do is to check the checkbox below the button to confirm the personality you entered. Don’t worry if your chatbot personality is not definite yet, you can always come back to this screen to make adjustments.

After clicking the Continue button, you get to the final setup screen. Here you can select which channels your chatbot will be active on after it goes live. Simply select the channels you connected earlier and click the Finish button.

Congratulations! You have successfully created your chatbot! But the work is not done yet. If you want your chatbot to actually help your customers, you will have to provide it with knowledge. Time to give this chatbot its brain!

Adding knowledge

Now that your chatbot has been set up, it’s time to provide it with the knowledge to successfully help your customers. To turn your chatbot into a conversational superstar, navigate to the Chatbot menu in the top menu bar. You can now see five steps for building your chatbot, the first of which is titled Add knowledge. Clicking this step shows that there are two ways to add knowledge to your chatbot: Conversations and FAQ.

watermelon dashboard chatbot builder view

Conversations

As the name suggests, with the conversations you can create entire dialogues for your customers to have with your chatbot. Customers can start a conversation by asking a question to which the chatbot can reply and even ask follow up questions or perform follow up actions.

To create a conversation, click the Add a new conversation button and enter a title for your conversation. After clicking Save, you get taken to the conversation builder screen. Here you can build your conversation, step by step. On the left side of the screen you can find the modules you can add to conversations. The modules are divided into three groups: triggers, actions and responses.

Triggers are forms of input from customers, used to start a specific conversation. In Watermelon there are two types of triggers: the start button and question variants. The start button is the first button customers can click inside the website widget on your website. When customers click the button, the conversation you’ve associated the button with will automatically start. Note that you can only use the start button once in your chatbot, so make sure to use it in the welcoming conversation of your chatbot.

If you make use of Artificial Intelligence in your chatbot, your customers can also trigger specific conversations in your chatbot by typing in questions. In the three grey fields, at the top of your new conversation, you can fill in the question you want to answer in this conversation, along with at least two variations on this question. These variations are used to provide your chatbot with context for the question and will help your chatbot be able recognise hundreds of other variations on this same question.

With the variants filled in, your chatbot will be able to recognise the customers’ questions, regardless of phrasing or spelling, match it to the correct conversation and give the right answer. To learn more about Artificial Intelligence for your chatbot, read this complete guide to AI.

After you’ve selected the trigger for your conversation, you can see that the Action modules are now available. These modules signify actions your chatbot can take in the conversation. For example, you can use the Text module to have your chatbot send a message to your customers. Attachments like pictures, gifs or files can be sent using the Attachments module. And when you’ve created multiple conversations, you can connect them to each other by using the Link to conversation module. Of course at the end of your conversation you can finish up with the End of conversation module.

When you have at least one Action module in your conversation, the Response modules become available as well. Responses are ways your customers can interact with the chatbot during the conversation. After the chatbot has asked a follow up question, you can use the Response modules to create multiple choice buttons your customers can click. Or you can let your customers fill in an input field or one of your custom fields. Other response options include a Calendar module for selecting dates, Checkbox answers for selections from a list or a Feedback module for asking your customers to rate their chatbot experience.

In your enthusiasm for building conversations, it can be easy to lose track of what you’ve built so far. To get a quick overview of your entire conversation, click the Open tubemap button in the top right corner. This opens a schematic flowchart of your conversation so far. Simply click on the modules to navigate to them within your conversation.

For more tips and tricks on building your conversations, read this article.

Frequently Asked Questions

The second way of adding knowledge to your chatbot is using FAQ’s. Where conversations are used for making complex dialogues with follow up questions and actions, you can use FAQ's to quickly answer questions using the AI of your chatbot.

To add your questions, select the FAQ menu and start by adding a new category. You can now click the Add new questions button to enter your first FAQ. Here you see three grey fields for your question and an input field on the right for your answer.

The grey fields are used for the question you want to answer, and at least two different variations on that question. Variations are different ways your customers might phrase the same question. For example, a question like “when do you open?” could have “What times are you open?” as a variant.

By adding at least two variations to your question, your chatbot can automatically learn to recognise hundreds of different variations, including spelling mistakes your customers might make. Of course you can add more variations by clicking the plus icon. We do recommend keeping your variants at a maximum of five or six per question. This way you minimise the chance of overlap and your chatbot will become more effective. To learn more about using AI in your chatbot, check out this guide.

With the question and the variants filled in, you can now type in the answer you want your chatbot to give. If you’re satisfied with the answer, you can click Save. After saving the answer you can see that an extra option has become available, the Forward option (the arrow icon). With the Forward option, you choose to have your chatbot automatically forward the conversation to an available agent or to a specific team.

In this article you can read more about adding FAQs to your chatbot.

Training & testing your chatbot

Once you’ve added all the relevant knowledge to your chatbot, it’s time to do some training. This is an important part of building your chatbot, as this is the moment when your chatbot will actually learn the knowledge you just added.

Navigate to the Chatbot menu and select the Training option. If you see a red warning sign next to the Training button, it means that new knowledge is added to the chatbot that hasn’t been trained yet. Luckily, training the chatbot is a completely automated process. You Just have to start it.

When you click the Train chatbot button, the chatbot will automatically start to analyse all of the new knowledge that has been added. Depending on how much knowledge you’ve added, this process can take two or three minutes to finish. So sit back, relax, or go grab a slice of Watermelon!

When your chatbot has finished training, you might be tempted to publish your chatbot to the world. But let’s not be hasty, as there is one more important step to take. To make sure your chatbot is completely ready to serve your customers, it's useful to do a round of testing.

In the Chatbot menu, select the Testing option to open up the internal testing environment. Here you can try out the chatbot conversations and FAQs before they go live. To start testing, simply click your Start button on the right and navigate through your chatbot as if you were one of your customers. If you encounter anything that needs changing, simply go back to one of the previous steps to update your chatbot.

Launching your chatbot

If you’re completely satisfied with your chatbot, the last thing to do is to activate it. This exciting step can be done by simply opening the Activate your chatbot menu and selecting if you want to activate your chatbot, suggestion bot, or both.

The suggestion bot is an extra tool available if you use Artificial Intelligence for your chatbot. Where your regular chatbot is there to talk to your customers, the suggestion bot has an internal focus to help your agents. During conversations your agents might have with your customers, the suggestion bot is there as well, reading along with your agents. The suggestion bot will analyse and recognise the questions your customers are asking, and will give your agents suggestions for answers to those questions. So even when the chat has been handed over to a human agent, your chatbot is there to help!

Once you’ve made your choice, click the big green button and your chatbot is ready to meet the world!

chabtot conversations watermelon dashboard

Step 4. Maintain success

Now that your chatbot is live, it’s important that your new digital colleague stays successful.

Next step

Boost Your Productivity with your own GPT-4 chatbot . Get Started Today.

Transform your Customer Service with Watermelon's AI and GPT-4 Powered Chatbot Solution. Try it free now.

Book a free demo