AI Makes Supply Chains More Resilient, But Only Up to a Point, New Study Finds

When the COVID-19 pandemic ground global trade to a halt, when geopolitical conflicts severed shipping routes, and when climate disasters flooded factories and ports, one uncomfortable truth became impossible to ignore: the world’s supply chains were far more fragile than anyone had assumed. For companies in developing economies, the reckoning was especially harsh. Now, a new study from Bangladesh offers some of the clearest evidence yet about how artificial intelligence can help firms in resource-constrained markets weather these storms, and, crucially, when it cannot. The research, published in Discover Artificial Intelligence, surveyed 372 managers across Bangladesh’s manufacturing and pharmaceutical industries and found that AI genuinely strengthens supply chain resilience, but that its power fades precisely when conditions become most turbulent.

The study, conducted by Md. Obaidul Hoque of Comilla University, focused on two of Bangladesh’s most economically vital and disruption-prone sectors. Manufacturing, particularly the export-oriented ready-made garment and textile industries, and pharmaceuticals together form the backbone of the national economy, yet both remain acutely vulnerable to transportation delays, supplier dependency, export market volatility, and climate-related risks. Despite this exposure, Bangladesh remains in the early stages of AI adoption, hampered by technological limitations, low digital competence, financial barriers, and limited organizational readiness. Many firms still rely on reactive, non-data-driven decision-making and traditional forecasting methods that perform poorly in high-uncertainty environments. This makes the country an ideal laboratory for testing whether AI’s celebrated benefits in developed economies translate to settings where digital infrastructure is thin and data quality is inconsistent.

At the heart of the research lies a theoretically grounded distinction between two concepts that are often conflated. Drawing on the Resource-Based View, the study treats artificial intelligence as a strategic technological resource, encompassing intelligent automation, machine learning tools, and algorithm-based systems. Predictive analytics, by contrast, is framed as an organizational capability: the managerial routines and competences that convert AI-generated data into forecasts, risk assessments, and operational decisions. This distinction matters because, according to the Resource-Based View, technology alone rarely confers competitive advantage. Value emerges only when resources are transformed into capabilities that are valuable, rare, and difficult to imitate. Contingency Theory then supplies the second analytical lens, holding that the effectiveness of any internal capability depends on the fit between the organization and its external environment. Together, these frameworks allow the study to ask not just whether AI builds resilience, but how, and under what conditions.

The methodology was rigorous for a developing-market survey. Data were collected between July and October 2025 through structured questionnaires administered both online and offline to supply chain managers, operations managers, procurement officers, logistics managers, and IT or data analytics professionals in the Dhaka and Chattogram divisions. Of 650 professionals invited, 400 completed the questionnaire and 372 responses were deemed valid, a response rate of 57.2 percent. All constructs were measured on five-point Likert scales using items adapted from previously validated studies, translated into Bangla through a back-translation procedure with two bilingual experts. The analysis employed partial least squares structural equation modeling in SmartPLS 4, with bootstrapping based on 5,000 resamples, and the sample size exceeded the minimum requirements indicated by both the inverse square root and gamma-exponential methods.

The results paint a nuanced picture. AI showed a positive and significant effect on predictive analytics, with a path coefficient of 0.36, and a smaller but still significant direct effect on sustainable supply chain resilience, with a coefficient of 0.17. Predictive analytics itself exerted a strong positive influence on resilience, also at 0.36, and partially mediated the relationship between AI and resilience with an indirect effect of 0.13. All effects were significant at the p < 0.001 level. In practical terms, this means AI strengthens supply chains both directly, by improving visibility, coordination, and responsiveness, and indirectly, by equipping firms with the analytical capability to anticipate disruptions before they escalate. The partial nature of the mediation also suggests that other organizational capabilities, such as digital integration, organizational learning, and supply chain collaboration, likely transmit additional portions of AI’s influence.

Perhaps the most striking finding concerns environmental uncertainty. Rather than amplifying AI’s benefits, as some earlier research had suggested, uncertainty significantly weakened them, with a negative moderating effect of −0.13. The explanation lies in the mechanics of machine learning itself. AI systems depend on consistent historical data and recognizable patterns; when markets swing unpredictably, regulations shift abruptly, or climate events shatter established operating rhythms, the assumptions underpinning algorithmic forecasts break down. In Bangladesh, where digital infrastructure is limited, supply chain networks are fragmented, data quality is inconsistent, and skilled AI professionals are scarce, these institutional barriers become even more severe during turbulent periods. Environmental uncertainty thus acts as a boundary condition, constraining the resilience benefits that firms can extract from their AI investments.

The study’s multi-group analysis added a sectoral twist. After confirming measurement invariance across the two industries using the MICOM procedure, the researchers found that the core relationships, from AI to predictive analytics, from AI to resilience, and from predictive analytics to resilience, held in both manufacturing and pharmaceutical firms. However, the moderating effect of environmental uncertainty differed sharply between them. In manufacturing, uncertainty significantly weakened the AI-resilience link, with a coefficient of −0.22, whereas in pharmaceuticals the effect was negligible and statistically non-significant at −0.01. The between-group difference of −0.209 was itself significant. The authors attribute this asymmetry to the pharmaceutical sector’s greater regulatory compliance, higher digital maturity, and more standardized operational processes, which appear to buffer its AI systems against environmental turbulence. Manufacturing firms, exposed to greater market fluctuations and operational complexity, are correspondingly more sensitive.

The model’s explanatory power tells its own story. AI and its downstream effects accounted for 58 percent of the variance in sustainable supply chain resilience, a moderate-to-substantial figure, but only 13 percent of the variance in predictive analytics. This gap suggests that AI alone does not fully explain how analytical capabilities develop; factors such as data quality, digital infrastructure, analytical skills, organizational readiness, management support, and digital maturity likely play important roles that future research should capture. The authors also urge caution in interpreting overall model fit, noting that the estimated model’s SRMR value of 0.174 and NFI values below 0.90 do not meet conventional benchmarks, and that the cross-sectional, single-informant design limits causal inference.

For managers in emerging markets, the practical message is clear: buying AI is not the same as benefiting from it. Firms must pair technological adoption with strong internal analytical capabilities, investing in data quality management, integrating data across supply chain partners, and cultivating the skills needed to translate algorithmic outputs into real-time decisions. In volatile environments, AI should be deployed alongside operational flexibility, diversified supplier networks, collaborative information sharing, and contingency planning. The sectoral findings add a further layer: manufacturing managers, whose operations are more exposed to environmental turbulence, should prioritize risk monitoring, supply chain visibility, and adaptive planning when implementing AI-powered systems.

The policy implications extend beyond the factory floor. Governments in developing economies, the study argues, should accelerate digital transformation by investing in infrastructure, promoting secure data-sharing platforms, and offering incentives such as tax breaks, matching grants, or subsidized data management systems for firms that adopt standardized data practices. Sector-specific data standards and interoperability guidelines could encourage common formats across supply chain partners, while partnerships between governments, universities, and professional training institutes could help alleviate the shortage of qualified AI and analytics workers. Because manufacturing firms face greater exposure to environmental uncertainty, policymakers may need tailored support mechanisms, including targeted digital innovation incentives and supply chain risk management programs. Ultimately, the study reframes AI not as a standalone shield against disruption but as a resource whose value depends on the organizational capabilities built around it and the environmental conditions in which it operates, a conclusion that resonates far beyond Bangladesh’s borders.

Subject of Research: The role of artificial intelligence and predictive analytics in building sustainable supply chain resilience under environmental uncertainty in Bangladesh's manufacturing and pharmaceutical sectors

Article Title: Artificial intelligence in sustainable supply chain resilience with predictive analytics as a mediator and environmental uncertainty as a moderator

Article References: Hoque, M. O. (2026). Artificial intelligence in sustainable supply chain resilience with predictive analytics as a mediator and environmental uncertainty as a moderator. Discover Artificial Intelligence, 6(1), Article 1365. https://doi.org/10.1007/s44163-026-02314-9

Image Credits: AI Generated

DOI: 10.1007/s44163-026-02314-9

Keywords: artificial intelligence, supply chain resilience, predictive analytics, environmental uncertainty, Bangladesh, manufacturing, pharmaceuticals, PLS-SEM, Resource-Based View, Contingency Theory, developing economies, sustainability