Your approach toward chatbot visitor questions and other input 

A big part of what makes a chatbot good is its ability to understand what the end user is asking. Before you can begin to create your organization’s bot, including the answers it provides, you must define how the chatbot’s answer is triggered.

In this article, we cover the following related topics: 

• Open vs. closed questions 
• Defining the expected end user questions and input 
• What type of bot suits your organization’s purposes best?
• Practical advice

Open vs. closed questions 

How would you like your bot to start the chat? We suggest two different options to begin the conversation between your chatbot and its users: open and closed questions.

Open questions

By asking an open question such as "How can I help you?", you’re letting the end user decide where to lead the chat. In this case, the chatbot visitors are given the freedom to type anything and select the first conversational topic.

To understand the visitor's freely written input, you must draw from Certainly’s AI tool, which is based on Natural Language Understanding (NLU). NLU is the AI discipline dedicated to understanding what humans say or write. 

The bot’s AI can be trained to understand any end user Intent as long as it is provided with enough data (the more data, the better). Over time, the bot can become more innovative, basing responses on customer preferences and providing recommendations. You'll learn much more about this AI feature throughout your chatbot project journey!

Closed questions

By presenting the user with a closed question, you’re only giving them a couple of suggestions to choose from. Those options are provided in the form of buttons the end user can click on. For example, you can display a list of popular topics or frequently asked questions.

We call these option buttons Suggested Replies. This method of starting a conversation allows the end user to easily navigate between topics the chatbot definitely has an answer for.

If you use buttons for the entire conversation, this is called a click-based chat. In this case, the end user will only ever choose between the predefined Suggested Replies given by the bot.

Our suggestion: A hybrid solution

Consider using both approaches. For instance, you may begin with a closed question but enable your chatbot’s AI at the same time. Over time, you'll collect enough data to train the bot's AI, and it will learn how your end users navigate through these conversations. Once you are confident enough, you can enable the open question at the start.

Ultimately, how to start the conversation depends on your specific use case. Nevertheless, we recommend that the bot is always set to understand freely written text.

Defining the expected end user questions and input 

Now it's time to figure out exactly what type of input you want your bot to be able to answer. In doing so, you need to gather a list of end user questions or requests that will lead to a chatbot answer that provides value. This content can be, for example, your most frequently asked or product-specific questions.

It’s always a good idea to ask your coworkers who regularly communicate with customers to help with this process. Many bot builders opt for an internal brainstorm session with their front-end colleagues to understand what kind of end user behavior can be expected.

This planning and input can be very valuable as you define the framework of your organization’s chatbot. 

What type of bot suits your organization’s purposes best? 

Here, we identify four bot types and the parts of the customer journey they relate to:

To plan what content should be used for whichever bot type you decide on, anticipate during which part of the customer journey the bot must provide support. For example, if you create a Basic Support Bot, you should add questions from your FAQ to the list and ask your customer support team what kinds of questions they usually receive.

You can read more about setting the right expectations, including the specific purposes of each of the chatbot types identified in the table above and their related business department activities. This will help you further refine the chatbot's answers, as you can ask for expert knowledge and content checks.

Practical advice

• When making your list of questions, try to put yourself into the end user’s shoes to anticipate what they might ask the bot.  
• You may draw from historic chat conversations of end users and your organization's customer support, or look into requests sent to your general support email.  
• Ask front-line support agents and sales personnel for their help. After all, the people who have the most contact with end users will know the most about what questions these customers have!

To summarize, when deciding what you want the bot to answer, determine what type of bot suits your organization’s purposes best. Those purposes must be aligned with the context or framework your organization sustains with its website.  

Consider whether you want open or closed questions...or perhaps a combination! And finally, get advice from the people around you with the most customer experience, as their input can make this process much more manageable for you.