The Defensive Equilibrium - Balancing between Algorithmic Efficiency and Human-Centric Sustainability

Sammanfattning

The integration of Artificial Intelligence (AI) into corporate sustainability promises to alleviate the cognitive boundaries of the Strategic Decision-Making Process (SDMP) by synthesising massive, complex datasets. However, the inherent friction between the reductionist, instrumental logic of algorithmic models and the pluralistic, value-laden logic of sustainability creates profound operational and ethical paradoxes. Employing a qualitative, multi-case study of the Swedish manufacturing sector and utilising the Gioia methodology, this thesis investigates how corporate decision-makers navigate the tension between algorithmic efficiency and human control. The findings reveal a stark literature-practice gap: rather than experiencing seamless "agentic delegation," decision-makers face severe socio-technical friction driven by fragmented sustainability data architectures. Consequently, managers experience extreme epistemic uncertainty and actively engage in "strategic non-use." They assign AI to administrative synthesis while retaining strict, defensive human authorisation over high-stakes sustainability choices. This study conceptualises this phenomenon as the Defensive Equilibrium, a path-dependent state where locally rational risk-aversion, combined with domain-neutral IT governance, collectively traps the firm in a state of systemic algorithmic under-use. To dismantle this equilibrium, the thesis proposes a Human-Centric AI Governance Approach that maps domain-specific ethical guardrails and continuous "algorithmic interrogation" protocols directly onto the phase-gates of the SDMP. Ultimately, this research argues that the strategic value of AI in sustainability is entirely contingent upon resolving foundational data constraints and formally institutionalising human moral accountability.