Jaunākie Biznesa Informātikas studentu zinātniskie raksti (publikācija daļēji angļu valodā)

Ir iznākuši trīs jauni esošo BI studentu un absolventu zinātniskie raksti. Raksti ir publicēti CEUR Workshop Proceedings: Joint Proceedings of the BIR 2025 Workshops and Doctoral Consortium co-located with 24th International Conference on Perspectives in Business Informatics Research (BIR 2025).

Toward a knowledge management method for training customer support AI agents

Edgars Dzenuska, Peteris Rudzajs

This paper summarizes preliminary findings on a knowledge management method to train generative AI agents for customer support in software companies. Despite advances allowing AI deployment with minimal technical skills, companies struggle with documenting and maintaining suitable knowledge bases. A survey of 20 software firms found that 75% face challenges in training AI with domain-specific knowledge. Through literature review and ndustry analysis, this research develops guidelines for creating and managing knowledge articles as training data. We report initial results from a ten‑article pilot in a European software company, where our method improved answer quality, as measured by BERTScore F1, Cosine similarity and human‑rated correctness of answers.

Pilns raksts pieejams te!

Addressing functionality gaps, data integrity, and system interoperability in enterprise systems

Matiss Gaigals

Mistakes and errors are an integral part of almost any complex enterprise system. Many of them face challenges, including limited functionality, data integrity issues, and interoperability problems. As a result, user operations may be blocked, requiring manual fixes. This paper proposes a method designed to detect such issues and provide AI assisted temporary mitigations with user oversight. It has been implemented and evaluated by an experimental application, with the open-source code made publicly available. The method has been validated within three practical scenarios, demonstrating its effectiveness in handling tactical failures in enterprise systems.

Pilns raksts pieejams te!

Structuring complexity: A functional evaluation of Jira tools for requirements management

Agnese Rozenberga

In complex software environments, effective requirements management (RM) is essential for aligning stakeholder needs, managing uncertainty, and maintaining traceability across development stages. While platforms like Jira are widely used, they lack comprehensive native support for core RM functions, leading to reliance on third-party extensions. This paper presents a structured, lifecycle-oriented method for evaluating and selecting RM tools within the Jira ecosystem. Building on established RM functional dimensions, a decision support tool was developed to help users identify the tools best suited to their project-specific needs. The tool enables users to prioritize RM capabilities and compare them against actual tool performance across six lifecycle stages: elicitation, analysis, specification, validation, management, and traceability. Expert feedback validated the relevance and usability of the tool. The findings underscore the importance of structured tool selection in managing RM complexity, especially in environments where no single solution offers full lifecycle support. The study contributes a replicable approach to RM tool evaluation and offers practical guidance for improving decision-making in tool adoption and configuration.

Pilns raksts pieejams te!