How to Design Chatbots More Elegantly

What’s the difference between basic button-based automated bots to NLP-driven conversational chatbots? Most importantly, how should we design a conversational bot that meets business needs? Here we discuss it together.

Different Types of Chatbots
First, it is important to distinguish between the various types of chatbots available in the market. From simple menu/button based chatbots to conversational AI chatbots, are there any levels? There are different types of chatbots and different technologies used, so let’s see what are the characteristics of each of them.

Button/menu based chatbot
As the name suggests, this type of chatbot lets users choose from multiple options, which are presented in the form of menus or buttons. Depending on what the user clicks, the bot will prompt him with another set of options to choose from, and so on.

As you can guess, they are structured in very basic form of buttons (single-choice, multiple-choice, etc.), and because of its simplicity, they represent most chatbots. These bots can answer pre-set questions and help users navigate a website or online store to facilitate their buying journey, but they are less effective when addressing complex requests involving a large number of variables. In fact, once the user’s query is not within the pre-set range, this type of chatbot can’t provide any assistance and end up making the user very disappointed and frustrated.

Keyword-based chatbot
With this type of chatbot, the user enters a word or a phrase and the bot recognizes the keywords in the query. This type of bot uses a basic analysis engine to process these keywords and match them to a preloaded thesaurus.

The advantage of this is that the bot will only reply to content manually loaded into the system and will not go off topic, allowing the business to have a friendly control over the automated messaging of the thesaurus.

On the other hand, such chatbots are limited by their inability to recognize misspelled words or slang. They are also highly contextual, and when used outside of their context can be significantly underwhelming. Ask the library chatbot a question about “book a hotel” and it might return books about hotels.
NLP-based conversational chatbot
This type is by far the most advanced AI chatbot. They use artificial intelligence and natural language processing to provide users with the best experience. Thanks to these techniques, the bot considers the different words that make up a sentence, analyzing them along with any available context to gain a contextual understanding of the question. So it can apply that understanding to the resolution of the query.

The main advantage of conversational chatbots using NLP is that they understand the meaning behind the words, and the algorithm-based advantage is also able to understand misspellings, thus providing a better user experience for the user.

Types of Answers Presented at Each Level of Conversational Chatbots
It is believed that every enterprise has already used chatbots with conversational AI technology, and the capabilities of robots can also reach different “levels” of conversations. Let’s take a specific case as an example to explain what these different stages look like.

Suppose a company develops an NLP conversational chatbot for internal use to answer employee questions about various HR matters. A team member wanted to know how many days of annual leave he had left, he asked the chatbot.

The first level of answer lies in telling the employee where he can find the answer to that question, usually on his payroll or HR software. This is the simplest and most basic level of conversation that can be easily achieved when designing a conversational chatbot.

The second-level answer is slightly more evolved, as the robot can redirect the employee to a specific internal system, such as HR software in this case, where he can find out how many days of annual leave he has left.

Finally, more advanced third-level answers allow the chatbot to automatically and seamlessly log the employee into the HR software so that he can directly access the information he needs. A bot at this stage can even prompt employees to apply for some annual leave via a calendar or form without leaving the chat platform. This stage obviously means that conversational chatbots can integrate with third-party platforms or software to be able to retrieve information into another system. This is one of the technical prerequisites for robots to provide such interactions and services.

How to Design a Smarter Conversational Chatbot
Having a conversational chatbot that uses NLP technology is a great start and can give your company a competitive advantage and cost-effectiveness, but you also have to make sure that the interaction with the bot is qualitative and meaningful to you attractive users. So how do you design a bot that users will reason to talk to? Here are some tips and previous practice examples.

Transaction based query script
As the name suggests, a script for a chatbot is a scenario that uses a predesigned conversational message (business process) as a response to a user query. Of course not all queries will require scripts: simple FAQ-type questions will be answered with a one-time request, but transactional queries will require scripts. In practice, bots have to follow a specific conversational flow to gather the details needed to provide specific information, such as a car insurance quote bot previously developed at an insurance company.

The flow will obviously present different prices depending on the chatbot input and the number of vehicles, but keep the following tips in mind when writing the flow:

The goal of the chatbot should be clear, it is best to achieve only one goal in a process
Keep the bot’s answers short and clear
The bot communicates as clearly as possible
During the session with the user, when there are unclear questions, try to guide the user with guiding words
Clarify your bot goals
Whatever the goal of your conversational chatbot, you have to make sure people understand it. This means that every response from the bot must be clear and free of any ambiguity that could lead to misunderstandings.

This may seem obvious, but most companies or Botmasters forget this simple rule. It results in a very confusing and unusable conversational interface, which completely defeats the purpose of designing the bot in the first place.

In addition to designing clear and well-defined processes, we must also keep the bot’s answers as short as possible. The reason is simple: the more you read, the more likely users are to become confused, tired, and distracted. A good way to do this is to break up the conversation, i.e. break the bot’s message into smaller chunks.

personalise
Personalization is your robot’s strength. In fact, we have to define what kind of personality we want the conversational chatbot to have in order to determine its tone, what language it will use, how it will communicate, etc.

Designing a user-friendly character is a tricky problem. Give it too little personality and the interaction feels tedious. Overuse it can quickly become annoying…

To sum up, designing a high-quality conversational chatbot is not an easy task, but I hope these tips and hands-on experience will be helpful when designing intelligent bots.