Dos and Don’ts of Designing Chatbot Conversations

Customer support has evolved over the years from the traditional call center to omnichannel support. Digitalization has made customers expect brands to have an extensive online presence. Yet, most support teams have a tough time managing the ever-flowing requests from various channels. While they are busy answering their queries, a significant portion of users leave your website without asking their questions.

Though it’s not humanly possible for your support team to reach out to customers every single time, there is one way to get such users to initiate a conversation. It’s the chatbot present on your website and social channels.

According to a report by Grand View Research, the global chatbot market is expected to reach $1.23 billion by 2025, an annual growth rate of 24.3 percent. With 45 percent of users preferring chatbots as the primary mode of communication for customer service inquiries, brands have turned towards artificial intelligence (AI) and machine learning (ML) for implementing chatbots.

And, one of the main reasons for this transition is the chatbot’s ability to give instant responses, which is a hit among all types of customers.

Let’s start with the basics — What’s a Chatbot?

In definition, a chatbot is a computer program that chats with humans and helps resolve their queries. Until the arrival of AI and ML, making the computer understand human language was a mammoth task. Currently being developed with natural language processing (NLP) algorithm, this technology enables the chatbot to understand the intent of users and answer accordingly. More of it in a later post.

bot for business

Though chatbots show a lot of promise for improving customer support, if they are not sufficiently trained, they will only make things worse for your support team. One of the most common instances of failed chatbots is the mismatch between the user’s query and the chatbot’s reply.  This could be due to gaps in the NLP algorithm resulting in a chatbot having trouble understanding what the customer is asking. Irrelevant answers from the chatbot further frustrate the customer.

However, if you take the necessary steps you can turn chatbot conversations around in your favor. Here are the dos and don’ts which help make your chatbot customer-friendly.

Don’ts First

Chatbot Dont's

Don’t Build a Chatbot that Takes Hours to Reply

The very purpose of installing a chatbot on the website is to provide instant replies. It wouldn’t make sense if the customer opens a conversation with a chatbot to find ‘Typically replies within a day’. On the other hand, a chatbot’s prompt and instant reply has the potential to drive your customers towards your brand. Take a look at Hugs chatbot which falls short of impressing users (despite its cute intro) due to its delayed response time.

chatbox replies

Don’t Make Your Chatbot Provide All the Support

While chatbots are meant to be a helping hand for your support agents, it wouldn’t be a good idea to have them take over all support interactions. Leave the monotonous queries for the chatbot to answer so that your support agents can take care of the more complex support issues. This is because customers prefer to interact with agents regarding important aspects like pricing and product customizations. In such situations, have your chatbot make an escalation to live agent as it is always better to have a live agent answer complex questions than risking an irrelevant response.

Don’t Robotize Your Chatbot

With personalization hitting even cold emails, it is only fair that your chatbot has a personality. Design your bot to converse in a style and manner that matches your buyer personas. In case you don’t have a very serious audience, this is the time to use your creativity. For example, there’s this website which has made an animal version out of its chatbot with light-hearted dialogues that bring a smile. This polar bear a.k.a chatbot named UX Bear teaches UX-related topics entirely through conversations.

Chatbox UX

Don’t Ask for Sensitive Information

Enabling payment through your chatbot makes it appear smart but comes with a few drawbacks. Many social platforms like Facebook and Telegram have provisions for integrating payments but not without safety risks.

On the other hand, most customers hesitate to reveal their private details such as credit card information and social security number to a chatbot. Considering the data breaches that take place frequently, it’s better to keep such requests for sensitive information to your app or website.

Chatbox sensitive info

Now for the Dos

Chatbox Dos

Do: Have an Option for Push Notifications

When customers have multiple tabs to attend to, there is a high chance that they might miss your chatbot’s reply and drop out of your website. Delivering the replies in the form of push notifications can decrease the number of customers leaving your site to a great extent.

Push notifications let the customer know that there’s a reply from your website even when they are on a different tab. Make sure to encourage your customers to ‘allow push notifications’.

And, never force them to give their email address to begin a conversation with your chatbot. The customer would rather leave your site than talk to your chatbot. Check out how Freshdesk chatbot requests for push notifications and email address from the user.

freshdesk chatbot
Be subtle in your approach with customers!

Do: Ask for Feedback

The technology around chatbots is still developing and you need to be all ears for feedback. Every small improvisation takes your chatbot one step closer towards understanding the customer. While feedback forms are completely okay for email support, it’s best to keep a miniaturized feedback for a chatbot. Ask your customers to rate whether they found the chatbot’s solution useful or not. This lets you find the areas where your chatbot needs improvement in terms of customer experience.

Do: Structure and Optimize Your Knowledge Base

The way of framing a question varies with each person and this can confuse a chatbot when it doesn’t match the questions in your knowledge base. This problem can be solved by structuring the knowledge base in a way that is bot-friendly. This will allow the chatbot to scan the content effectively and pull up relevant solution articles for the customer. Such content structuring increases the probability of understanding the questions more accurately. So, make sure that you are constantly feeding it necessary information.

Freshdesk knowledge base


Brands need to be very careful when designing their respective chatbots. A well-designed chatbot is a big boost to your brand reputation and customer acquisition. On the flip side, one impersonal remark is all it takes to drive the customer away from your brand. Building a smart chatbot can be quite tricky. Follow these tips mentioned in the article to make sure that your chatbot offers instant assistance in every possible way. After all, a smart chatbot is not built in a day!

Try Freshdesk

Leave a Reply

Your email address will not be published. Required fields are marked *