The Governance Gap Behind AI Deployment in High-Informality Economies: Board Oversight, Procurement Asymmetry, and Operational Risk

Contemporary artificial intelligence governance frameworks have focused predominantly on downstream concerns such as explainability, bias detection, transparency, robustness, model performance, safety, and post-deployment monitoring. Yet a more fundamental institutional problem remains insufficiently examined: whether organizations possess the governance capacity, operational visibility, procurement structures, institutional accountability mechanisms, and board-level oversight architecture required to responsibly deploy artificial intelligence systems before deployment occurs. This practitioner working paper examines the governance deployment gap, defined as the divergence between successful technical deployment and the institutional conditions required for sustainable, accountable, and operationally resilient AI adoption. The central argument is that deployment success can conceal governance fragility when organizations are capable of operationalizing AI systems despite lacking the institutional structures necessary to supervise, challenge, independently evaluate, or sustainably govern those systems over time. This paper argues that many governance failures do not originate exclusively from defective outputs, biased models, hallucinations, or post-deployment incidents. Instead, governance vulnerabilities frequently emerge much earlier, during procurement, deployment decision-making, institutional implementation, vendor dependency formation, and organizational adoption processes. In these contexts, technically functional systems may operate within structurally fragile governance environments characterized by asymmetrical information between vendors and institutions, limited internal technical capacity, fragmented accountability structures, weak procurement oversight, insufficient board visibility, operational dependence on external providers, and reduced institutional autonomy over critical technological infrastructure. Existing governance discussions often assume that deployment itself constitutes evidence of institutional readiness. However, organizations may successfully operationalize AI systems while lacking the governance maturity required to understand deployment dependencies, evaluate long-term operational risks, supervise external vendors, maintain internal auditability, or sustain institutional accountability over time. As a result, technical deployment can create the appearance of institutional competence even when the underlying governance architecture remains incomplete, fragmented, or operationally weak. Focusing on high-informality economies in Latin America, this paper examines how these governance conditions become especially significant in environments where uneven institutional capacity, fragmented digital infrastructure, procurement asymmetries, limited oversight resources, operational urgency, and dependence on external technology providers frequently coexist. In such contexts, organizations may deploy advanced AI systems into environments where documentation systems, institutional coordination, procurement supervision, and operational governance mechanisms remain unevenly developed. This paper further argues that governance fragility may remain institutionally invisible precisely because deployment appears operationally successful. Systems may continue functioning, scaling, generating outputs, and supporting operational processes while underlying governance weaknesses accumulate through vendor concentration, procurement opacity, insufficient institutional challenge capacity, fragmented oversight responsibilities, weak escalation pathways, and limited internal understanding of system dependencies. In this sense, operational functionality does not necessarily indicate governance resilience. To operationalize this governance problem, this paper proposes the Institutional AI Deployment Review (IADR), a structured pre-deployment governance review architecture designed for boards, executive leadership teams, procurement functions, institutional oversight bodies, and large organizations evaluating AI adoption at scale. The IADR is not presented as an entirely new theory of governance, nor does it claim to invent new governance dimensions. Its contribution lies in synthesizing established principles from the NIST AI RMF, ISO/IEC 42001, OECD AI Principles, NACD governance guidance, and WEF governance discussions into a unified deployment readiness architecture specifically adapted to AI deployment conditions in complex institutional environments. The contribution of this paper lies in reframing AI governance from a predominantly model-centered problem toward an institutional deployment capacity problem. Rather than centering governance exclusively on model explainability or output fairness, this paper shifts attention toward institutional preparedness itself: whether organizations possess the structural conditions required to govern deployment before it occurs. This includes examining procurement asymmetries, institutional dependency formation, oversight fragmentation, escalation mechanisms, board visibility, operational resilience, vendor governance, accountability allocation, and institutional continuity under conditions of technological dependency. Deployment governance challenges are amplified in high-informality environments where institutional systems frequently interact with fragmented records, uneven administrative infrastructure, constrained technical capacity, and operational pressures that incentivize rapid adoption over governance consolidation. In these environments, governance capacity and deployment capacity often evolve at different speeds, creating conditions in which organizations may scale AI adoption faster than their internal governance structures can mature. This paper does not present empirical causal claims or statistical inference. It does not claim to demonstrate universal causal relationships between deployment structures and governance outcomes. Instead, this practitioner working paper advances a conceptual and institutional governance framework intended to support board-level oversight, procurement review, deployment governance, operational supervision, and institutional decision-making in environments where AI deployment capacity and governance capacity do not necessarily evolve in parallel. The original contribution of this paper lies in identifying and structuring a governance problem that remains underexamined in much of the current AI governance literature: the possibility that institutions may become operationally dependent on AI systems before developing the governance architecture required to responsibly supervise them. The governance deployment gap describes not merely a technical problem, but an institutional asymmetry between the speed of deployment and the slower development of governance maturity, operational accountability, and institutional oversight capacity. This paper complements the author's prior work on pre-evaluative exclusion (Velarde, 2026, DOI: 10.5281/zenodo.19665795), which examined how algorithmic systems may exclude populations before evaluation occurs. Together, the two papers articulate institutional asymmetries on both sides of evaluative AI deployment in Latin America: one concerning who or what becomes admissible to evaluation before assessment occurs, and the other concerning whether organizations possess sufficient governance integration capacity to supervise operational AI dependency once deployment occurs.