Behind the scenes: Developing the PremFina’s AI Chatbot. - Questers

Behind the scenes: Developing the PremFina’s AI Chatbot.

The recently announced chatbot “Fina” of our partner PremFina has been a huge success for them. The bot has reduced the load on their customer services team with 77.5% for the last two months and continues to improve. PremFina’s team at Questers was the main driver of the development of the bot and therefore, we talked with our colleague Hristomir Todorov, a BackEnd Developer, to shed some light on the chatbot and how he was involved in the development process. Read what he shared:

1. You were actively involved in the development of PremFina’s new chatbot. Can you share some details about it? What is its main purpose?

As our business grows and we get more clients, the need for more client support increases. Our goal is to automate everything, and the next logical step was to automate the most frequently requested tasks from our customer service team. Therefore, Fina, the bot, was born.

2. You started developing the chatbot in April and two months later it was live on the official PremFina website. Tell us more about the development process and the tech stack you’ve used?

We have put much work into developing the chatbot. It all started with a 3-day virtual hackathon, in which we decided what technologies to use to develop the chatbot. The next two months we spent developing and polishing the functionalities, making it easy and fun to use. With the help of our customer service team, we added dialogue and trained the bot to understand different customer intents.

3. Can you share more about your personal involvement in the project?

I was involved in the development of the bot from day one with Kamen. Most of my work involved creating the base technology for Fina so that our users are able to interact with our platform in a natural way. We tried different approaches and helped train Fina accordingly to make the customer experience fun and easy.

4. Did you have any bumps in the road? Can you share what was the most challenging part of this project?

Getting Fina to integrate with our software was easy and quick. The challenging part was to combine the different overlapping intents that Fina has to distinguish as an AI tool and perform the respective functionalities accordingly.

Directing our users to customer service team presented another challenge. We chose a new chat platform for our advisors, and we had to integrate it with Fina. The result is a seamless transition for the client from Fina to our customer service team.

5. Which part of your work are you mostly proud of?

I am mostly proud of the way Fina interacts with clients. It is all chat-based and seamless. I have seen many bots that communicate mostly with buttons, and they function as a knowledge base. Fina tries to understand what the client really wants and offers answers or actions. The natural language processing of the bot is continuously improving, and every day our customer services team is looking into the history and training Fina further to improve its knowledge.

6. Are you planning to add some new functionalities to the Fina chatbot?

We are currently gathering the requirements for the next significant feature of Fina, and development should start very soon. It will be complex and will require a lot of checks and validations, but we are always ready to challenge ourselves.

7. Any other interesting projects you are currently working on?

There are a lot of fun and exciting projects in PremFina. We are using the latest cutting-edge technologies, and every project is impressive. There are other ongoing projects in the AI field as well and they are scheduled to go live in the near future.


Read our insightful interview with Kamen Kalchev, a BackEnd developer at PremFina’s team at Questers, who was also heavily involved in the development of Fina chatbot.

Also, make sure to check out the hot vacancies we have for the Sofia team of PremFina. Their team is growing and we are looking for more proactive and motivated people to join them.

field_tagged: