As AI agents become central to organisational efficiency, their integration into specialised knowledge work domains such as HR project management remains underexplored. At Endress+Hauser, a global leader in measurement and automation technologies, HR project managers face recurring challenges including administrative overload, governance complexity, and difficulties maintaining strategic alignment, yet no AI-supported tooling exists to help them navigate these workflows. Existing research addresses AI in HR or in project management separately, but not in their intersection, and the role of Explainable Artificial Intelligence (XAI) in building trust in this combined domain has not been empirically examined. This thesis addresses that gap.

The research follows a Design Science Research (DSR) methodology guided by the FEDS framework (Venable et al., 2012), structured across five phases. The Awareness phase combined a literature review with expert interviews with five HR project managers and one portfolio lead, identifying six operational challenges and seven AI-integration risk categories, with participants unanimously treating XAI as a prerequisite for adoption. The Suggestion phase translated these findings into 16 design requirements. The Development phase produced Hanna, an AI agent configured in Microsoft 365 Copilot Agent Builder and grounded in nine internal Endress+Hauser documents, with ex-ante validation by two domain experts. The Evaluation phase assessed Hanna with the same HR stakeholders across six realistic scenarios using a mixed-methods design combining Likert ratings with open-ended qualitative feedback.Hanna was evaluated positively overall across the three target dimensions, which are efficiency, strategic alignment and trust with explainable AI. Time saving averaged M = 4.17/5, and the strategic alignment scenario averaged M = 4.50. Source transparency (M = 4.17) and reasoning transparency criterion (M = 4.07) both averaged above 4.0, while trust in the output averaged 3.60, a consistent gap identified across all six scenarios. This finding, that perceived explainability does not automatically translate into output trust, is absent from existing XAI and HR project management (PM) literature and represents the central theoretical contribution of this thesis. Eleven recommendations for the next development iteration are proposed, primarily through instruction set refinements.

Finally, the conclusion answered the research questions.The thesis confirms that AI agents can be successfully integrated into HR PM at Endress+Hauser to enhance efficiency and strategic alignment, where XAI plays an important role in fostering trust. Practically, Hanna serves as a direct proof of concept for the full Copilot Studio deployment planned for Q3, and the DSR evaluation design used here offers a reusable template for future enterprise AI agent studies.