Al in Startup Evaluation: Revisiting Benefits, Risks, and Decision Bottlenecks in Volatile Economies

The growing adoption of artificial intelligence (AI) in startup evaluation has generated strong expectations regarding improved investment decisions. However, the majority of existing studies implicitly assume stable economic environments, transferable data structures, and predictive accuracy as the primary indicator of effectiveness. This systematic review revisits these assumptions by examining the role of AI in startup evaluation within economically volatile and institutionally constrained contexts. Drawing on an in-depth analysis of several academic studies, the article addresses three core questions: (i) the feasibility of deploying AI-based evaluation models in unstable innovation ecosystems, (ii) the principal advantages and limitations of such models, and (iii) the structural and behavioral risks associated with their real-world implementation. The review reveals that while AI models are capable of extracting informative signals from heterogeneous startup data, their standalone predictive performance is fundamentally constrained by data bias, contextual non-transferability, and irreducible uncertainty. The findings suggest that the primary value of AI does not lie in replacing human judgment, but in augmenting decision-making through structured risk analysis, cost-sensitive reasoning, and scenario-based insight. Building on this synthesis, the article proposes an uncertainty-aware, human-in-the-loop Decision Support System (DSS) architecture that embeds AI as a cognitive-risk analytical layer rather than a definitive decision authority. By shifting the focus from prediction accuracy toward decision architecture design, this review contributes to the entrepreneurial finance and decision support literature and offers practical guidance for investors and policymakers operating in volatile startup ecosystems.

کلیدواژه ها:

Artificial Intelligence ، Decision Support System ، Structural Risk ، Cognitive Decision-Shaping