[INTERVIEW] The Challenge Behind Creating a Chatbot – a CSM Story

Updated on November 1, 2021

Back in the day, a customer service chatbot was nothing more than a fancy toy. Its vocabulary was extremely limited when responding to something that was beyond its reach. Availability has changed with the Internet development, and so have the user expectations. Artificial intelligence is progressing rapidly, outperforming humans at nearly every cognitive task. As a result, chatbots became capable of maintaining adequate conversations, more or less, based on dictionary, context, and specific syntaxes.

Bots are perfect for Customer Success, giving the user an interactive experience. If functional, bots can help customers find what they are looking for, and help resolve their issue, without any human interaction whatsoever.

A screenshot of Sales.Rocks chatbot

But the process of developing a company chatbot is not that easy. It has its ups and downs. Ana Ilievska, a Customer Success Manager in Sales.Rocks has experienced many of them. In this article, she explains her journey for creating a user-friendly chatbot she named SalesRocky.

Ana, as a Customer Success Manager, why is it important for a business to have a Customer Service chatbot?

First of all, it is important for us, to make our jobs easier. We are alive human beings, and we cannot be online 24/7. If you take into consideration the different time zones in the world – some customers/visitors might be in the middle of their working hours when sales agents are asleep. So, having a bot that is always awake and ready to help, comes very handy for Customer Success Representatives.

Also, there are people that are introverts and don’t want to talk to a live agent. For them, it is easier to talk with a robot.

Can you elaborate on the process of developing Rocky?

I made the customer service chatbot with the programs that Zoho offers. They give you a choice of different platforms, Dialog Flow being part of them. I started using one of them by entering the content about each and every question a person can ask for our platform, about the modules and how to use them, pricing, credits, etc.  After I entered maybe half of the questions, I decided to test it before finishing it completely. That was the moment when I noticed that something isn’t right. It turned out that this platform was not working very well with word recognition, which is important for understanding the content context before replying.

So, instead of changing it to Dialog Flow, I continued working with it and entered the questions and answers with all possible varieties, hoping I would fix it. In the end, there were around 1600 questions, but the bot was still not working. (sad face)

We also wanted to enable a Single Selection Option. For example, when somebody visits our website, Rocky would pop up and present a few common questions for them to choose from, which is a perfect option for people who are visiting our website for the first time. This is the way to give them an idea of what they might need to know about our platform. Unfortunately, the first platform didn’t give us that option, so finally I switched to Dialog Flow.

It turned out that Dialog Flow works a lot better for us. It functions on word recognition, which means that it recognizes everything that you ask, and I didn’t even enter half of the questions like before. The Single Selection Option is also available. This was done by setting up a Dialog Flow Fulfillment Code that will provide users with suggested answers that might help them easily decide what they are looking for. Now, I am proud to say that the chatbot is finally working, and Rocky is smart at last. (smiling)

A screenshot of Sales.Rocks chatbot - Single Selection Option

Basically, creating a chatbot means knowing everything about your products, setting up a database with questions and answers, and writing programs for including additional options, in our case – the Single Selection Option.

What was the biggest challenge while creating the chatbot?

There were 2 main challenges. The first one was getting to know every single thing about the Sales.Rocks’ platform, which was essential in order to make Rocky capable of automatically replying to website visitors.

For this, I had help from our product owner Olgica. I shadowed her work daily, and had many questions, eager to learn everything. The second challenge was working with Zoho and integrating their programs with our needs, especially when it came to making and fixing the Single Selection Option. The Fulfillment code we needed, plus the configuration between Sales IQ and Google Dialog Flow were a bit problematic, but in the end we figured it out together with Zoho.

How did you decide on the name SalesRocky?

The name of the platform is Sales.Rocks so I wanted it to be connected with Sales and Rocks. My first idea was Sally. There was a suggestion from my coworkers that the name should begin with Sales, but SalesSally didn’t sound nice, and requires some serious tongue twisting to pronounce it right. (smiling) I believe that the bot should be nice and friendly to everyone that needs help, so he needed to have a friendly and catchy name (yes, it’s a he), and that is when I came up with SalesRocky.

Author avatar