In 2025, organizations scramble for early generative AI (GenAI) adoption, often with mixed-at-best results. This tends to be diagnosed as some combination of overhyped technology and incompetent implementation. I offer a different explanation, namely, a fundamental mismatch between the GenAI capabilities and the legacy design logic of human-centric organizations.
The talk will frame organizations broadly as sets of institutional rules and choice environments designed to (1) support the alignment of human workers with the organizational mission, and (2) mitigate or harness human capability limitations (e.g., cognitive biases). GenAI also displays a range of alignment failures and capability limitations. However, these do not overlap neatly with their human counterparts. Therefore, releasing GenAIs into the human-centric institutional ecosystem introduces a foreign element that often interacts with the legacy organisational logic in suboptimal or destructive ways.
This happens both in the direction of GenAIs mitigating some expected human limitations (e.g., enabling new forms of performance metrics gaming) and GenAIs catalyzing and compounding human limitations through their own. I will sketch several organisational failure modes and ponder the prospects of AI-resilient, AI-empowered, and AI-native institutions.
This talk is presented by the Technology and Sustainable Futures research group and the Sustainability Lab. Coffee and refreshments will be served, and after a 45-minutes presentation, we’ll have ample time for questions and discussion.