What is a conversation?

In my last post, I wrote about how the conversation is the message. This post covers how your bot needs to understand the conversational context and respond appropriately so that your brand has a successful conversation with your end-users. 

There are some great pieces and advice on conversational architecture: greeting, on-boarding, fallback, ending, keeping messages short and simple, using quick replies and carousel cards. However, just as important is the conversation itself and what makes or breaks a conversation. The measures you can take to avoid a failed conversation by understanding context, conversational versatility, and using different tones of your brand’s voice will be my main focus.


Conversation as a process is interactive communication between two parties where questions might be asked and answered, opinions made, sentiments offered, observations stated, and ideas exchanged. For this to work, the two parties need to share a common language that is both spoken and understood by them.

Besides sharing a language, there needs to be shared knowledge and a basic understanding in terms of content and context for any conversation to make sense. By defining and limiting expectations and predicting behaviors, we simplify the process of encoding meaning and ensure the flow of communication.

Let’s take the following example; if you asked me about the weather and I replied that my favorite pizza topping is pepperoni, our conversation would leave you feeling confused and lost for words and even maybe a little frustrated. Through experience, you have learned that to get an answer; you need to ask a question. Whether the person can or can’t answer you is beside the point. You expect to get several answers ranging from an actual statement about the weather, speculation or a simple I don’t know from me because it will still make sense to you and you know how to react to it naturally. However, my statement about pepperoni was an unexpected behavior that made no sense and left you with the feeling of confusion or even a little frustration and indecisive about how to react.

We learn through experience to expect communication to follow certain rules for us to be able to make sense and for the conversation to take a natural flow. It is where context comes in.


Conversation depends significantly on context. The most excellent conversationalists are those who know how to interact in any context.

Examples of different contextual factors to consider in a conversation.

So what is context? According to the online Oxford dictionary, it is the circumstances that form the setting for the conversation you’re having. When determining the context in a particular instance, you need to consider physical, historical, social, psychological, and cultural aspects. As an example of how physical context plays a role, let’s take expressing joy at a music concert and expressing joy in a church. Jumping around and whipping your arms in the air while screaming “woohoo” communicates a very different message at a concert than it would in a church. An example of the importance of psychological context is how moods and emotions impact what is said or how it’s understood. If I’m pissed off, even a genuinely heartfelt compliment can be understood the wrong way. I’d like to think I’m not alone in this (social context: if you knew me, psychological context: it’s meant to be funny).

Historical context is very relevant for a bot in terms of what was said and shared in past conversations. A bot that can’t consider historical factors often leads to bad user experience. Take, for example, when an end-user return and the bot introduces itself all over again (not remembering the user). Or asks the user over and over again for the same information to direct them to a particular path.

If your bot doesn’t understand the context, your end-users will be left feeling misunderstood, ignored, or, in the worst case, insulted.
It doesn’t mean that just by giving the correct answer to the question, the bot is a great conversationalist. Your bot is a good conversationalist when it can react appropriately in different situations. For example, after asking a user how they’re doing, you can get responses ranging from “I’m on top of the world” to “life sucks” and the bot has to pick up on the psychological context and reply appropriately. It means being more empathetic with users that are sad before firing the next question — asking users if they want to speak to a human colleague or asking if the bot can help if that’s relevant for the use case. Of course, the bot has a job to get done and needs to take the user down the main path of conversation as it was designed to. However, you also want to avoid ignoring emotional cues shared by your end-users at all costs.


It’s every conversational designer’s wish that most end-users play nice and go through the main conversational path created for the bot. However, experience proves otherwise and shows that there are many different types of users, and some people can’t help but poke at the bot and test its capabilities.

So you need to guesstimate user inquiries and responses and create conversational content around that. Let us say you’re creating a Customer Care bot and your colleagues have provided you with all the FAQs they receive. The main conversational path will be the bot’s responses to the FAQs. However, a good conversationalist should be able to answer other non-FAQ related questions about the weather, name, location, or even tell a joke. The bot doesn’t necessarily need to give the correct answer. An answer that shows it understands that a question was asked and in return, gives a somewhat sensible response and gently directs back to the main path will offer a good user experience.

If your bot sticks to the main path and doesn’t allow for little nuances in the conversation, it’ll appear to lack minimal conversational skills. It will offer a less than mediocre user-experience. So being aware of building in responses to questions or comments that might fall out of your golden path of conversation will be worth the effort in the long run.

Brand Voice and Tone

So a bot with excellent conversational skills should be versatile and offer the right content depending on the context. We, humans, do that by changing our tone of voice and using our body language. However, as most chatbots are of a textual format, a bot can only use words and emojis to convey the tone of voice.

A bot that can impress users with its conversational skills is a bot that can string words together in a manner that fits the tone of voice for that particular situation. So if your bot picks up that a user is feeling frustrated, your bot needs to respond with a choice of words that convey that it’s understood the user’s frustration and help the situation by providing the best possible answer or ask if the user would prefer to speak to a human colleague. The choice of words and the tone your bot uses needs to compare with your brand’s tone. Because the bot is your brand talking to your end-users, the way the bot makes users feel translates into how your brand makes them feel. It’s an important thought to always keep in mind when writing responses.

I’m going to delve deeper into brand voice and creating a tone spectrum in my next post because of its importance and simply because it would make this post too long to read. In my opinion, defining the voice and tone is the first and most significant step in creating the bot’s persona. So keep an eye out for that!

Stay tuned for more!


Want to know more about Chatbots – Book a Demo!