Artificial Intelligence Improved the Efficiency of MND's Customer Center Operations

MND Energie, a.s. supplies gas and electricity across the Czech Republic to more than 250,000 customers, both households and businesses. Alongside electricity supply, it also offers the Solar Account, which allows customers to make the most of their home photovoltaic systems. It purchases all surplus electricity without limits and at a price known in advance. The company has further expanded its offering with the Flexibility service, through which customers can earn additional income by smartly managing their battery and electricity overflows, with a guaranteed minimum reward.
The company is part of the MND Group, the largest Czech company focused on oil and natural gas exploration and production, drilling and service activities, as well as the storage and trading of natural gas and electricity. In addition to the Czech Republic, MND also operates in, for example, Ukraine, Germany, Hungary, and Italy.

MND (Moravské naftové doly) processes a high volume of incoming email correspondence in its customer center (call center), reaching 300 to 500 emails per day (approximately 7,000 emails per month). These emails had to be read and manually sorted into more than 40 different categories (e.g., complaints, self-readings, etc.) so that correctly categorized emails would reach the CRM system.

This mechanical and monotonous work was performed by an average of 4 part-time employees (equivalent to approximately 2 full-time equivalents, or FTEs). The process showed high employee turnover, delays in handling (caused by limited availability of human capacity outside peak hours or on weekends), and a natural error rate caused by the human factor.

The team of AI specialists from Aricoma deployed a robust cloud solution built on the Microsoft Azure platform. At the core of the entire system is an advanced classification model within Azure Language Studio, which Aricoma's experts cleaned, trained, and "wrapped" in a containerized service running directly in the customer's tenant.

How the system works in practice:

• Preprocessing: Emails are downloaded in batches several times a day into a database, where the system carefully analyzes the content, that is, the subject and body of the email.
• AI classification: Based on the content, the language model determines the exact business category. This also includes an intelligent anti-spam filter that reliably detects spam, while using keywords to rescue legitimate emails with an advertising footer from being deleted.
• Automated execution (RPA): Because the legacy ERP system could not be easily integrated via API, UiPath software robotization was deployed. This software robot simulates user behavior and, based on the AI model's decisions, automatically categorizes requests in the ERP system.

The solution has been in operation for more than 2 years and shows high reliability. In December 2025, the model underwent planned retraining on a new data sample (100 to 300 examples for each category), which further significantly increased its accuracy. This AI solution concept is easily replicable for other customer centers of similar service providers.

The deployment of AI and robotization has brought MND clear business benefits. Full automation of routine email sorting has saved 2 full-time equivalents (FTEs). Call center employees can now focus entirely on resolving complex customer requests and on higher-value activities.

In the long term, routine and mechanical email sorting is an extremely monotonous activity, comparable to assembly-line work, which naturally leads to high employee turnover in call centers. By fully automating this agenda, the customer has eliminated its dependence on the human factor and removed the management burden associated with constant recruitment and training of new staff.

With manual sorting into more than 40 categories, classification was often subjective: what one operator assigned to one category, another could place elsewhere. The AI model has delivered consistent classification free of emotion and subjective bias, which has made data clearer across all clients and established a uniform standard for handling their requests.

Unlike human capacity, which is subject to turnover, sick leave, or limited working hours, the AI model operates as a service continuously and reliably. Dependence on the immediate availability of operators has been removed, which has reduced variability in email processing speed. Thanks to immediate savings on labor costs and the elimination of costs associated with turnover and training of new employees, the overall return on investment of the solution is very fast.