Upcoming Webinar: How government and public services orgs are using chatbots for citizen support









Register now
All Blogs

4 Best Practices For Conversational Design

April 08, 2020  •  3 minute read

Building a Virtual Assistant that can answer user queries and a Virtual Assistant that can serve as well as delight customers are two different things. It goes beyond just having a good model (though it most certainly helps).

Here are a few best practices of conversational design from the on-field observations of what makes a really good virtual assistant/chatbot.

4 best practices to build a chatbot

Step 1: Having a chatbot persona

Virtual Assistants in today’s world have become a really good investment for the companies and individuals who want to serve their customers 24×7

The boom has largely been due to the fact that chat and voice channels are much easier to access and interact with than the traditional mode of communication.

The point to keep in mind though is that these new channels seem to be easier because they feel informal, less restrictive, more fluid, and above anything else, more personal.

Hence, it becomes paramount to provide that personal touch.

The users should feel that the assistant is very approachable and they can ask it anything (gracefully handle the queries which are out of scope). A good step is keeping the tone polite and courteous. This helps users feel at ease. 

Another good step is to think of the assistant as a real person. Therefore, create a whole personality around it. Let the bot have some quirks and anecdotes, keeping in mind the brand image that you are trying to portray.

Step 2: Making the UI beautiful

As people generally judge other people on how they look in the real world, the same is true for the virtual world. If the virtual assistant doesn’t wow them with the intuitive way it can represent information, or showcase its quirks, it quickly bores the end-users.

Remember that people are coming to the assistant to easily get and digest information. A text message may be presenting the whole information, but showing it as a card or some images would really help the end-user to pursue it better. After all, humans understand and retain information more easily in visual form.

Step 3: Making the UX intuitive

This point ties into the previous point. It’s one thing to have beautiful graphics/UI and another thing to bombard users with it.

Remember that anything in too much quantity just saturates the experience. UX comes into play as a component that helps you figure out what is the best way to maintain the experience. It provides you with a way to know whether having a card or a quick reply button will be better for the type of information.

Whether the information being presented is able to grab the attention and help the user.

 Step 4: Virtual Assistant’s Knowledge Base

Coming to the last, but not the least point, the Assistant would only be as good as the Knowledge Base it is equipped with. Now, you may ask, that this is the most obvious thing, right? Well, there are many ways to look at it.

The first and the most important principle to follow is, 

Garbage IN = Garbage OUT”

What does this mean is that if you feed garbage to the underlying model, you can’t expect the end result to be world-class? Focus on the quality rather than the quantity of the knowledge being fed and you will find a sharp jump in the quality of response that the assistant can provide.

So, as the 80/20 rule goes, focus on the 20% data that 80% of your users are most concerned about, increase the quality of knowledge that the assistant has about these scenarios, and this, in turn, ensures a high customer satisfaction (CSAT) score for the assistant. It takes painful observations and iterations to generate a good quality knowledge base, but the end result is far superior to anything else.

Remember, a good assistant is akin to Alfred or Jarvis, they can help you out on the tasks so that you can increase your productivity on the things that matter.

Siddharth Goel

Read more
Build your first Dynamic AI Agent in under 10 clicks