What is conversational design?

Conversational design ensures that a conversation between people and artificial intelligence is possible as casually as possible. The aim is to make this dialogue between people and systems feel natural. Particular emphasis is placed on the fact that interaction, for example with a chatbot or a virtual voice assistant, which is perceived as helpful and pleasant.

The basis for this is human conversation and an understanding of human interaction. Ideally, a conversation with a chatbot looks like a conversation with a person. Nevertheless, a chatbot should never pretend to be human, as this would lead to misunderstandings and false expectations. In addition, conversational design also deals with human psychology so that systems learn how people think and behave in a conversation.

Chatbots need conversational design

Without a well-thought-out conversational design, a conversation with a chatbot quickly looks wooden or is not very helpful — and that starts with the greeting between chatbot and customer. Because only with an appropriate conversational design does a chatbot also greet and say goodbye to a person, which they would otherwise not do at all. And even the greeting is important, because the chatbot always represents the company and so the language of the chatbot should also suit the target group.

With a good conversational design, dialogs with a chatbot feel natural and intuitive. This not only provides better user experiences, but also increases efficiency, increases customer satisfaction and saves costs because fewer personnel are required for repetitive tasks:

• ‍User experience: A good chatbot always focuses on the user's needs. The intentions are quickly identified and relevant and targeted information is provided in a user-friendly manner. The better the conversational design of a chatbot, the better the user experience.

• ‍efficiency: A good conversational design ensures that inquiries are processed quickly and efficiently. It helps to reduce processing time, minimize misunderstandings, and automate repetitive tasks. With a clear structure in dialog design, users can communicate their concerns more easily and quickly, because precise and understandable wording leads to fewer misunderstandings. This also increases the initial resolution rate, the direct problem solution after the first contact.

• Customer satisfaction: Customer satisfaction also depends on the conversational design of a chatbot. When the bot quickly and easily offers a solution or the desired information, users are happy. Conversational Design should therefore offer clearly understandable and precise dialogues so that the conversation feels natural. The better the interaction, the happier the customer — and this in turn strengthens the bond with the company.

Do's & Don'ts Conversational Design

When it comes to conversational design, companies can quickly step foot in fat. So that doesn't happen, here are the most important tips:

✅ Put on customer glasses when designing answers

The answers should be formulated in such a way that the user's perspective is really reflected. It is important to use only necessary information. At the same time, seemingly simple aspects that seem commonplace for the company should be explained and broken down in detail. This prevents misunderstandings. It is also advisable to precisely define the customer's expectations. What can the chatbot clarify and what not?

✅ Appropriate intro provides clarity

At the beginning of an answer, an introduction from the chatbot that fits the topic can provide a better understanding. A sentence such as:”I am happy to help with the following topics:...”. If the chatbot is unsure which topic the question belongs to, it can play out a request. This is necessary, for example, if spelling errors or dialects are found in the request. In this way, the chatbot ensures that the request is correctly understood and ensures even smoother communication.

✅ Several small blocks of text for answers

In general, answers should always be kept as short and concise as possible. If the answer is a bit more detailed, it makes sense to divide the text into several small blocks of text and play them out piece by piece. This makes readability easier, especially on mobile devices, and slightly loosens up the text. It is the same as with endlessly long WhatsApp messages, which only very few people are happy about.

✅ Offer multiple choice questions  

By using multiple choice questions, individual answer paths can be played out to the customer. With the help of buttons, the user can easily select the appropriate option, which speeds up the process and improves the user experience. If more detailed processing is necessary, the chatbot can forward the customer to a customer service employee, who will then take care of the request.

❌ Use more than 3 slides for image or video material as an answer

In addition to plain text messages, images or videos can also be used as response content. These are particularly advantageous for step-by-step instructions or portfolios. They can also lighten up a text a bit. It makes sense not to use more than three sliders, as the click rate decreases after the third slide.

❌ No monotonous answers  

It should be avoided that all answers sound the same. It is therefore important that the content of the answers is variable and formulated individually, using language that is as natural as possible. In addition, technical terms should be avoided in order to make it easier for the user to understand. In addition, it is useful to include information about the next step in the answer so that there are no uncertainties.

Using conversational design to find the right chatbot for the target group

In addition to technical development, conversational design also plays an important role in developing a chatbot for the target group. First, a target group analysis is carried out to find out which target groups use the chatbot and what characteristics it should have. Because even the chatbot can have a personality and this must match the brand. To achieve this, it is recommended to use so-called personas. Although this process is time-consuming, it is worthwhile in order to really understand the characteristics and needs of the target group. In addition, existing channels, such as telephone or live chat, can be evaluated to find out how users communicate, what frequently asked questions and which answers are particularly helpful.

The personality of the chatbot

The personality of a chatbot also reflects the identity of the brand, because as part of customer service, it represents the company. The chatbot should therefore adopt the company's values and tonality in its own language. There is a term for that botsanality, composed of the words chatbot and Personality.

Since the chatbot has a dual role as a mouthpiece and brand ambassador at the same time, it is advisable to actively shape the character of the chatbot. Often, the initial contact with the brand is taking place between chatbot and user. And this user experience also has an impact on the perception of the entire company.

Whether the personality of the chatbot should be formal or informal is determined by the company and the target group. In the leisure sector, people may use more informal language and use formal language in a traditional company. Different cultures and age groups also prefer a different approach. In principle, the tonality should be authentic and match the company's values and brand identity.

An example: There are three different types of chatbot personas: the avatar, the digital assistant and the “we brand”. The former is a fictional character, the second is the most neutral and the last is the most formal. Choosing the right chatbot persona always depends on the type of company. You can find out more about this topic in this article read.

Conversational design for chatbots: 5 things to consider 


1. Balancing language, technology and psychology

To understand communication between chatbots and humans, two different types of brains must be considered. Put simply: the human and artificial “brain.” The latter is more of a neural network. They both have their own abilities and limits and different triggers to make them work. While the artificial “brain” needs structured data and intentions as well as variables, the human brain needs empathy, guidance, and encouragement. Language helps both brains function and communicate with each other. Conversational design therefore requires technology, language and psychology to be in harmony.

2. Conversational design should build trust

First of all, communication with artificial intelligence, such as a chatbot, is a bit strange and can quickly drift off into the bizarre — especially if the conversational design is not consistent. However, with the help of a well-designed bot personality, trust can be built so that it feels as though communication is taking place with a real person. However, it is important to make it clear that this is a chatbot so as not to create false expectations. The goal is not to mislead the user, but to create pleasant and flowing communication.

3. People at the center of conversational design

People should always be at the center of the conversation and therefore also a certain amount of empathy. This allows the chatbot to better consider a user's fears, motivations, and potential context. This makes the user feel understood and is more likely to be guided to find a solution to their concerns. However, it is advisable to refer to a person when it comes to emotionally charged and health issues, as chatbots are more used to resolve recurring inquiries.

4. Use role-playing games

Role-playing games or sample dialogs help to simulate as natural a conversation as possible between humans and chatbots. One option is improvisational theatre to find out what makes a conversation human. One person plays the user and the other the chatbot. The conversation is conducted without visual contact so that only words are used to explain things. After a few runs, this allows you to find out the most natural course of the conversation.‍

5. The 80/20 principle

The chatbot does not have to know an answer to every question, but rather to provide sufficient information to the most common problems. The Pareto principle of 80/20 helps as a guide. 80% of users run through around 20% of the scenarios — namely the scenarios that are as simple and standardized as possible. For more complex issues, it is recommended to be referred to a person.

Conversational Design in Transition

Over time, the development of conversational design has changed quite a bit. In the beginning, the chatbot owners wrote manual dialogs for the chatbots and defined the answers themselves. Even today, this is still practiced in many cases, as GenAI is too risky for many companies due to hallucinations and possible fails.

And also when it is done with the help of generative AI or GenAI and their ability to learn from existing dialogs have made it possible to have new conversations without repeating themselves, not all companies are on-board. Many go by the motto: Better a standard statement than 100 different statements from 100 different customers.

Nevertheless, it is clear: With GenAI, it was possible to make communication between humans and chatbots more natural and break away from rigid patterns. Speed, flexibility and scalability in particular were positively influenced by this. With the help of generative AI, chatbots can be developed in less time and adapted for a wide range of use cases. GenAI not only speeds up the chatbot's response time, but also allows interactions to be automated. In addition, scalability is increased, as a large number of chatbot responses can be generated in different languages in a very short time. As a result, companies can serve more users with the same resources and significantly increase their reach.

Challenges and opportunities through automated text creation

Anyone who frequently works with generative AI is familiar with the challenges with quality, context and the legal and ethical statements in question. The texts often lack depth and creativity. They sound superficial and generic. In addition, checking the correct content is particularly important here. Certain requests may lack understanding of the full context.

With regard to ethical and legal statements, the texts created by AI must always be reviewed. Correction and control by humans therefore remain essential in order to deliver the desired quality and correct results. GenAI should definitely be taken with caution, as companies cannot afford to provide false information or mistakenly recommend products to competitors.

As a solution, moinAI offers its own GenAI with controlled frames so that hallucinations can be avoided. But answers created manually are also possible.

“Everything works for us, from manual to GenAI: Chatbot answers can be created manually or generated automatically. The spectrum ranges from completely manual to fully automated: Each answer can be entered by yourself, or the company — a digital assistant that helps with text creation and suggests answer concepts — provides support. Alternatively, GenAI generates responses based on specified sources. Either way, all the answers are in the moinAI Hub, which offers easy editing, clear statistics and an ideal UX — perfect for cross-team collaboration.”

On the other hand, automated text creation also offers great opportunities in terms of time and cost savings and adaptability. Texts can be created within seconds and particularly large amounts of text remain scalable. The AI can translate texts into several languages without any problems. In addition, quality assurance can be promoted by automatic corrections.

Controlling GenAI responses

Modern chatbot providers make it possible to influence the tone of voice of the texts, the use of emojis, etc. This allows you to define the following aspects, among other things, despite automatically created texts:

• salutation
• Emoji usage
• Tone of voice: humorous, such as Yoda, Hamburg neighborhood slang or Bavarian dialect, etc.
• Individual commands such as: don't output URLs, don't start answers with yes or no

In addition, GenAI responses can be controlled indirectly via the data basis on which they are based. The way information is described there can influence the AI's response. You can also give the AI specific commands for topics, such as “Always point out that we have changed our opening hours.”

In the future, AI chatbot providers, such as moinAI, will be able to define editorially stored snippets that are played out after a generated answer, e.g. a follow-up question or a CTA URL button.

conclusion

Conversational design is essential for successful use of chatbots, as this is the only way to have a casual conversation with artificial intelligence. Communication with a chatbot should feel like a dialogue with a person — but without imitating them. With a well-thought-out conversational design, the user not only receives helpful tips and quick solutions, but also evaluates the company more positively. The reason: A chatbot is part of customer service. Well-designed dialogs create trust in order to find out the reasons behind the user's concerns. Generative AI in the form of automatic text creation, can create texts in seconds and translate them into different languages. In the end, however, you always need the person who does the fine-tuning and ensures correct results and quality assurance.

A well-designed conversational design therefore remains the key to the success of a chatbot. The human touch should always be in mind, as we don't want to have the feeling of communicating with a machine when talking to a chatbot. Instead, we want a targeted, natural and informal dialogue. Curious about a real, natural chat? Then try out our AI chatbot.