How can artificial intelligence go beyond answering questions to work with business data, prepare product documentation or support decision-making in logistics? These are the possibilities being explored through a collaboration between Allium and the Department of Informatics at the Faculty of Business and Economics of Mendel University in Brno. In a project supported by the Operational Programme Technologies and Applications for Competitiveness (OP TAK), the partners are developing two specialised AI agents for business information systems.
The project, entitled “Development of Agents Using Artificial Intelligence”, combines Allium’s experience in developing business applications with MENDELU’s expertise in artificial intelligence, data processing and intelligent systems. The project aims to deliver two digital solutions, Pimics AI Agent and DLD AI Agent, designed to extend the capabilities of the Pimics system and the Microsoft Dynamics 365 Business Central environment.
Unlike a conventional chatbot, an AI agent is designed not only to understand a request but also to retrieve relevant information and suggest or carry out subsequent actions. For example, it could process a customer email, check product availability in a business system and prepare the information needed for a sales transaction. The aim is to combine natural language understanding with specific, controllable operations using company data.
Companies with extensive product ranges work with technical documents, catalogues, price lists, images and supplier information. Pimics AI Agent is being developed to analyse these materials, convert unstructured information into product database records, add product attributes, check data consistency and generate product descriptions. Users will be able to perform selected operations using natural language instructions.
One specific development scenario involves the automated creation of product sheets and catalogue materials. The agent is intended to combine structured data, images and a predefined graphic template to generate documents such as product sheets in PDF format. A key requirement is to maintain the factual accuracy of the information and keep content processing separate from graphic design.
DLD AI Agent focuses on business communication and logistics. It is being designed to recognise requests in emails, identify products, retrieve information from Microsoft Dynamics 365 Business Central and suggest responses or product reservations. Other planned features include stock picking recommendations and the identification of potentially lost sales.
The solution relies on cooperation between individual software components. One component evaluates the request, another processes the necessary business data, and a subsequent component prepares a proposed action or response. This integration is what distinguishes the agent-based functions under development from a standalone language assistant.
Developing AI agents requires more than simply deploying an existing language model. The university team is involved in designing the solution architecture, interpreting user instructions, preparing datasets, selecting and validating AI algorithms, and evaluating them experimentally. Key considerations include accuracy, response time, computational requirements and processing reliability.
One of the technological challenges is the development of a specialised metalanguage and its interpreter, designed to translate natural language requests into operations on business data. This approach also addresses limitations on direct database access in the cloud-based Business Central environment. The knowledge gained could also be applied to the further development of intelligent business applications.
The project runs from October 2025 to March 2027 and is divided into three phases. The first phase focused on requirements analysis, the design of methods and the basic system architecture. The second, scheduled to run until the end of September 2026, includes further development of individual components, optimisation of instruction interpretation and training of agent functions. The final phase will involve integration, algorithm refinement and testing the solutions in their intended operating environment. The functions described in this article therefore represent the objectives of ongoing development rather than features of an already completed commercial product.
Allium contributes its knowledge of business processes, customer needs and Microsoft Dynamics technologies. MENDELU brings expertise in artificial intelligence, data analysis, experimental development and decision support. The partnership also creates opportunities for early-career researchers and students to become involved and for the knowledge gained to be incorporated into teaching.
The project team from the Department of Informatics at PEF MENDELU includes Prof. Tomáš Pitner, Prof. Radim Farana, Assoc. Prof. Oldřich Trenz, Dr Jana Dannhoferová, Ing. Michaela Veselá and Bc. Sofie Kryštofová. The university team is led by Assoc. Prof. Oldřich Trenz.
The joint development of Pimics AI Agent and DLD AI Agent demonstrates how academic expertise and business experience can contribute to a new generation of business systems in which artificial intelligence not only helps process information but also supports decision-making and automates everyday tasks.
The project “Development of Agents Using Artificial Intelligence”, registration number CZ.01.01.01/01/24_062/0007603, is co-funded by the European Union under the Operational Programme Technologies and Applications for Competitiveness (OP TAK), Application – Call II – Development of Digital Solutions.