Problem-solver vs problem-creator: a multi-layered paradox of AI in innovation failure

Purpose This paper aims to address the fundamental paradox of Artificial Intelligence (AI) in innovation, exploring its dual roles as both a problem-solver for human-led failures and a problem creator of new systemic risks. The paper constitutes a broad, multi-layered framework for understanding this apparent paradox of failure and how it might be resolved. Design/methodology/approach A literature review is synthesized from different research streams, including management science, systems engineering, human-computer interaction and socio-ethical studies. This paper introduces a conceptual model for AI in innovation failure that disentangles the phenomenon into four cascading layers: social, process, technical and socio-ethical. Findings The analysis reveals a cascading failure model initiated by social breakdowns (e.g. collaborative disconnects), amplified by process deficiencies (e.g. translational gaps), manifesting in technical fragility and culminating in socio-ethical crises. Crucially, the study identifies near failures not as liabilities but as strategic inflection points that, when managed via specific interventions, stimulate organizational resilience. Research limitations/implications As a conceptual framework, the model needs to be empirically verified. One key implication is a need for cross-level research exploring the dynamics of the problem-solver vs. problem-creator paradox within each layer and analyzing how near failures facilitate organizational learning based on empirical data. Originality/value This paper’s originality lies in synthesizing siloed academic disciplines through the unifying lens of the problem-solver vs. problem-creator paradox. It proposes an innovative and holistic framework for both academic research and practice to identify, mitigate and manage the complex, dual-sided risks of AI in innovation, moving beyond a traditional cost-benefit analysis to a dynamic model of paradoxical tensions.