Understanding Value in Data, Analytics, and AI Initiatives to Support Value-Based Decision-Making and Prioritization : A Case Study in a Software Operations Context

Sammanfattning

Data, analytics, and AI initiatives are increasingly expected to contribute to business value, yet organizations often struggle to determine which initiatives should be prioritized. This challenge arises because the value of such initiatives is frequently uncertain, indirect, delayed, and dependent on organizational conditions. This thesis examines how value in data, analytics, and AI initiatives can be understood and structured to support value-based decision-making, enabling organizations to compare, prioritize, and select initiatives with the greatest expected contribution to the business. The study is based on an abductive qualitative single-case study within the software operations function of a global technology company. Empirical data were collected through two iterative rounds of semi-structured interviews, internal documents, and a form-based weighting exercise. The data were analyzed through a theory-informed qualitative process inspired by the Gioia methodology. The findings show that value-based prioritization is supported by structuring value as a layered assessment rather than as a single ranking of initiatives. Stakeholders first need a shared way of reasoning about value and the conditions required for value creation. Based on the findings, the thesis develops a layered framework that first screens initiatives against binary preconditions for value creation, including goal integration, ownership of value realization, data readiness, target audience identification, and a plan allowing iterative value discovery. Initiatives that pass this screening are then evaluated through gradable dimensions related to value realization and value impact. These dimensions are weighted and mapped in a portfolio matrix that distinguishes different value profiles, such as quick small wins, reliable value creators, transformational initiatives, and lower-priority initiatives. The study contributes by showing how Multi-Criteria Decision Analysis (MCDA) logic can be adapted to data, analytics, and AI initiative contexts through layered evaluation, separation of value realization and value impact, and portfolio-oriented classification rather than ranking alone. It also highlights the contextual nature of weighting and the need to balance analytical completeness with practical usability. The framework supports more transparent and structured prioritization, while recognizing that weights, classifications, and portfolio balance remain context-dependent.