Abstract

The integration of artificial intelligence (AI) into project management has accelerated dramatically, introducing a fundamental privacy paradox: the same capabilities that introduce vulnerabilities through data aggregation, sensitive inference, and third-party processing, also offer privacy protection through automated compliance monitoring and anomaly detection. This dual role is compounded by lifecycle diversity. Predictive, adaptive, and hybrid lifecycle approaches create different governance rhythms, stakeholder communication patterns, and risk management structures, meaning privacy risks that are manageable through front-loaded planning in a predictive project may emerge as sprint-level decisions in an adaptive one or require entirely different mechanisms at hybrid project phase transitions. Yet no empirical research has examined how project managers perceive and navigate these privacy risks in practice. Three interconnected gaps define the problem space: no empirical study has examined how project managers perceive and navigate AI privacy risks across different lifecycle contexts (theoretical gap), no lifecycle-matched, evidence-based understanding of privacy risk challenges specific to AI tool types (practical gap), and PMBOK 8 (Project Management Institute, 2025) conceptualizes risk as a core performance domain, but does not address how AI privacy risks emerge and differ across lifecycle approaches (governance gap). This dissertation addresses these gaps through practitioner perspectives rather than framework prescription. Using Straussian Grounded Theory (SGT) methodology, this dissertation investigates these gaps through semi-structured interviews with project managers across three lifecycle contexts. Contingency theory (Donaldson, 2006; Lawrence & Lorsch, 1967), privacy calculus theory (Dinev & Hart, 2006; Li, 2012), and sociotechnical systems theory (Trist & Bamforth, 1951), serve as sensitizing concepts without predetermining what the data will reveal. Data will be analyzed through SGT’s three-step coding process (open, axial, and selective coding) using ATLAS.ti, continuing until theoretical saturation is reached. The expected contribution is an empirically grounded theory of AI privacy risk navigation emerging entirely from practitioner experience.

Download

Share

COinS