GPT as an additional communication channel for customers

To give customers more options to connect with them, ROSSMANN has introduced a chatbot on its site. And not just any... GPT based chatbot. He will answer correctly 81% of queries.

Chatbot Eda in numbers

81 %

queries that came to GPT
handled GPT correctly

69 %

queries that went to the virtual assistant solved by NLP

92 %

queries that were received by the NLP chatbot correctly

ROSSMANN, one of the largest companies on the pharmaceutical market, has almost 180 stores in the Czech Republic with a turnover of CZK 6.5 billion. In addition to the Czech Republic, it also operates in Germany, Spain, Turkey, Albania, Kosovo, Poland, Hungary and Azerbaijan. It is the first in the Czech drug market to introduce cashless self-service cash registers in its stores and has also developed a digital loyalty program in the ROSSMANN CLUB mobile application, which includes the possibility of online purchases.

Default Status

The aim was to offer the customer help even when the customer line cannot be reached. Therefore, the chatbot was the ideal solution to this situation. Before him, the customer could contact customer service by phone, email or on social networks. ROSSMANN recognizes that process improvement, innovation and quality customer service are key to success.

Service selection

A good experience with the parent company was of great importance when deciding on a solution from coworkers.ai Daktela. The advantage then was the openness of the sales department and the quality of the product offered. Thanks to the full compatibility between the coworkers platform and Daktela software, customer support staff can actively respond to queries transferred from the chatbot directly within the Daktela system.

“The key was how to build a dialogue on the search for outlets. This dialogue combines cities where we operate a maximum of 5 stores with larger agglomerations, where we have more stores (Prague, Brno, Ostrava, Pilsen). For this dialogue, it is also important to mention the so-called. “carousel”, where there are more stores in a given location, the customer has the opportunity to conveniently browse from left to right between stores using this element.”

Jiri Kočis

Head of Customer Service and Reception, ROSSMANN

The solution

ROSSMANN uses new technology “GPT knowledge addon” from coworkers.ai combined with a chatbot running on NLP. The latter allows you to upload documents, tables and web pages to the virtual assistant's database with information from which it should draw for even more accurate answers. Thanks to GPT, it is possible to achieve the relevance of responses up to 95%. The introduction of a chatbot with GPT integration took approximately 4 months. A frequently asked question by customers is the possibility of employment in a company that the chatbot can solve playfully. A minor problem arose when solving the scenario for asking for information about branches, since in some cities there are more branches and they needed to be differentiated. However, this was achieved thanks to a carousel that offers all possible branches in the given location and the user can simply click on the details of the branch that interests him.

Results

In total, the GPT chatbot has learned 150 queries, namely 35 in NLP and 115 in GPT. Completing additional queries is very simple, you just need to update the file already uploaded in the administration or upload a new one.

Chatbot
The chatbot received 2197 queries for December 2023, and of 69.2% solved our NLP and the rest 28.4% solved the new product coworkers.ai GPT Knowledge addon. The rest 2.4% enquiries were sent from the bot to the operators

GPT Knowledge addon
The new GPT Knowledge addon, which was one of the first companies to implement ROSSMANN, was able to successfully solve 80.6% questions that customers have had about him. GPT knowledge addon works on the principle of uploading documents, tables and web pages to the virtual assistant database. From these, he then draws various information that customers ask about.

How does it work? - difference between classic chatbot and GPT chatbot

In the chatbot you can have scenarios set for order status or e.g.: basic FAQ. But in order to be able to answer a large number of questions, just as many scenarios would have to be written to do so. GPT can respond to anything you give it a background for - whether it's information from an annual report, a product list, or a contact on a website. And all this quickly, easily and with an accuracy of up to 95%.