Have you ever become really desperate during a conversation with a chatbot? Unfortunately, this happens very often. Instead of quick and uncomplicated support, the user ends up with a chaotic conversation after which he feels just as lost as before.
In this article, we’ll give you practical tips on how you can do better when creating your own chatbot. Because interactive, result-oriented chatbot conversations are no coincidence, but a result of following certain rules.
1. Consider the components of a conversation
A consistent structure is important in order to provide orientation. Each conversation in customer service usually consists of three parts: greeting, problem-solving and goodbye.
- Greeting: At the beginning of the conversation, the participants greet each other, and the customer service employee introduces himself.
- Problem-solving: The further course of the conversation is structured. The customer describes the problem and then the employee guides him through the necessary steps to solve his problem.
- Saying Goodbye: If the problem is solved, the employee will ask for further open questions. If the problem cannot be solved, the customer will be redirected to the responsible contact person or informed about the necessary steps to take. Afterwards, both say goodbye.
You should use the same structure that is common for the conversation with a human employee, for the conversation guidelines you create for your chatbot.
2. Lead the conversation
Let’s take a closer look at the “Problem Solving” part. In order to have a natural conversation, it is important to know and use its basic components (question,
confirmation, information transfer, etc.). That way the user can be guided through the conversation in a structured way.
Each component can be used in slightly different ways. For example, there are different types of questions that can be asked in a conversation:
- Yes/no questions (1): The user is asked a question that he can answer with Yes or No. For example: “Are you a customer of ours already?
- Either/or questions (2): With this question type, the user can choose between two options. For example: “Do you want to buy sneakers or boots?
- Open questions (3): This type of question is a bit more complex. It is introduced by question words (what, why, how, etc.). “What kind of shoes are you looking for?
Question types 1 and 2 can be handled well by chatbots, as there are predefined options to answer. For further improvement, you can use buttons to enable faster communication.
Question type 3 is a bit more complex to process and should be used with caution. Answers are not restricted in any way and can be very extensive. With this question type, you should work with hints. For example: “What kind of shoes are you looking for? We have boots, ballerinas, sneakers and high heels.” By doing so, the user knows what options are available and can formulate his answer accordingly.
Generally, every time an input is made, you should either confirm it or tell the user that the input could not be processed and what to do next.
Chatbots, voice assistants and AI,
stay informed with the free Onlim newsletter.
3. Be personal and human
No user wants to communicate with a machine that seems strange and impersonal to him. Users should know that they are communicating with a chatbot, but the bot should behave as humanly as possible.
You can achieve this following these tips:
- Filler words, emojis and pauses: All three are crucial parts of a messenger conversation. Use filler words and emojis for your chatbot conversations and delay the bot’s response by a few moments.
- Variations: If you ask a person the same question several times, he will usually answer a little bit different each time. Use several different formulations for the same statement for your bot as well.
- Context: Also try to create context within the conversation by connecting your answers to the user’s previous information or questions.
- Audiovisual content: Pictures or videos will help to make the conversation more engaging. They make an otherwise text-based conversation much more vivid.
For example, if the user has previously written that their favourite colour is blue and they are now looking for T-shirts, you can include this information in the bot’s response.
One possible formulation would be: “Hmm, let me think about it. (pause) How about this shirt in your favourite colour? :)” + photo/video/link to the product. With this answer, you don’t just create context. You also use filler words, emojis, pauses and audiovisual content.