Abstract Biodiversity loss is a systemic crisis produced by tightly coupled ecological, social, and economic processes. Yet, global biodiversity governance under the Convention on Biological Diversity (CBD) has largely operated through technocratic policy pipelines, including discrete targets, linear implementation logics, and the assumption that scaling interventions will yield systemic change. The Kunming–Montreal Global Biodiversity Framework (GBF) seeks to break with this pattern by articulating a whole-of-system governance architecture spanning both government and society. This paper examines how and why, despite this ambition, early implementation is reproducing familiar forms of fragmentation and selective action. Drawing on computational text-as-data analysis and network analysis, I analyze 3,417 national biodiversity targets and 718 biodiversity pledges submitted by 1,086 state, sub-national, private, and civil-society actors under the CBD. The analysis shows that implementation is shaped by stable belief systems and dominant policy narratives that privilege biophysical conservation targets while marginalizing enabling conditions understood as governance capacities required to address indirect and underlying drivers of biodiversity loss. These capacities cluster around four interrelated domains. Discursive capacities to problematize indirect drivers and articulate transformative narratives remain weak, leading to convergence around “safe” conservation themes. Institutional capacities for cross-sectoral coordination, policy mainstreaming, and regulatory and incentive reform are underdeveloped, limiting the translation of ambition into structural change. Resource capacities, particularly financial and administrative capacities, are unevenly mobilized and allocated, with biodiversity finance disproportionately directed toward countries signaling ambition or reflecting historical power asymmetries rather than those facing the greatest ecological pressure and governance constraints. Finally, action capacities, including participatory, implementation, and adaptive learning capacities, are constrained, resulting in weak coupling between problem, policy, and political streams and systematic narrative misalignments across actor groups. Together, these findings demonstrate how technocratic policy logics persist even within governance frameworks explicitly designed to be transformative. The paper contributes to policy process research by showing how computational text-as-data and network analysis methods can be used to empirically diagnose discursive, institutional, resource, and action-related capacity gaps in large-scale environmental governance, and by identifying the conditions under which whole-of-system approaches risk reproducing business-as-usual implementation dynamics. --------- This presentation is part of the session “Beyond Technocratic Pipelines: Problems-in-Solutions in Environmental Policy Processes” at the Conference on Policy Processes Research, Bern, Switzerland. The session is convened by Dr. Van Thi Hai Nguyen, with co-conveners Julie Zähringer, Stefano Jud, Mialy Rann, Fernando Fernandez, and Quynh Nguyen. Contact: Dr. Van Thi Hai Nguyen (van.hai.nguyen@wyssacademy.org / thi.nguyen@unibe.ch)
Van Thi Hai Nguyen (Thu,) studied this question.