Sharing downstream sales data with a neutral AI provider cuts grocery forecasting errors by about 40% and smooths predictions, helping to tame the bullwhip effect. The gains are largest when AI complements human planners and firms invest in digital skills, leadership and inter-firm collaboration.

How to use Artificial Intelligence to smoothen the bullwhip effect in FMCG supply chains

Weisz, Eric Ryan · February 01, 2026 · Kollektionen Digitale Sammlungen (SLUB Dresden)

openalex quasi_experimental medium evidence 8/10 relevance Summary only summary available; pdf_status=paywall DOI Source PDF

Structured author observations

Linked only from stored provider relations; the raw author line above is never matched by name.

OpenAlex

Latest observation: August 24, 2026

Weisz, Eric Ryan provider ID

Pooling downstream demand data through a neutral AI provider and using joint AI forecasting reduces forecasting errors by roughly 40% and produces steadier predictions, especially when AI augments rather than replaces human planners within organizations that have digital skills, leadership commitment, and collaborative practices.

Citation observations

Cumulative provider counts captured on specific dates; providers are never combined.

0 cumulative citations

OpenAlex · Observed August 24, 2026

View corpus context

The bullwhip effect, defined as the amplification of demand variability as orders move upstream, remains a persistent challenge in the field of supply chain management. The consequences of this effect include inventory imbalances, service disruptions and financial losses, which arise as firms base their decisions on incomplete or distorted information. Advances in AI have the potential to mitigate the bullwhip effect by integrating multiple data sources and supporting collaborative decision-making. However, the management literature has not yet clarified how AI can be effectively deployed to address this problem. To close this gap, the present cumulative dissertation brings together three studies to answer the central question: how can AI help mitigate the bullwhip effect in the supply chain? The work is structured around three sub-research questions. The first study asks what organizational challenges firms face when implementing AI to reduce the bullwhip effect. The second question pertains to the extent to which AI will function as a complementary instrument to human decision-makers or as a substitute for prevailing forecasting and planning systems. The third study investigates how data collaboration and AI-driven joint forecasting can reduce the bullwhip effect. Paper I conducts a systematic literature review and proposes the Bullwhip-Smoothing-Framework (BSF), which highlights digital skills, leadership commitment and collaboration as essential management pillars. Paper II proposes an AI Collaboration Framework comprising five stages: complementary, augmentative, collaborative, autonomous and replacing. The paper explores how AI supports information sharing and trust and stresses the need for trust in AI technology. Paper III presents an empirical study of joint AI forecasting in a grocery retail supply chain, demonstrating that sharing downstream data with a neutral AI provider reduces forecasting errors by approximately forty percent and produces steadier predictions. Across the three studies, the results indicate that combining AI with organizational competencies (digital proficiency, leadership development and collaboration) can effectively reduce the bullwhip effect.It is also demonstrated that AI is most effective when it complements rather than replaces human expertise and when partners share downstream demand data to train joint models. The dissertation concludes by presenting theoretical contributions, practical recommendations, limitations and directions for future research.

Summary

Main Finding

Combining AI with organizational capabilities (digital skills, leadership commitment and interfirm collaboration) can substantially mitigate the bullwhip effect. AI is most effective as a complement to human expertise, and joint forecasting using shared downstream demand data—mediated by a neutral AI provider—can reduce forecasting errors by roughly 40% and produce steadier predictions upstream.

Key Points

Problem: The bullwhip effect amplifies demand variability upstream, creating inventory imbalances, service disruptions and financial losses driven by incomplete/distorted information.
Research question: How can AI help mitigate the bullwhip effect in supply chains?
Structure: Cumulative dissertation with three linked studies addressing (1) organizational challenges for AI implementation, (2) whether AI substitutes or complements human decision-makers, and (3) the effects of data collaboration and AI-driven joint forecasting.
Paper I: Systematic literature review → Bullwhip-Smoothing-Framework (BSF). Identifies three essential management pillars for AI success: digital skills, leadership commitment, and collaboration.
Paper II: Conceptual AI Collaboration Framework with five stages of human–AI interaction: complementary, augmentative, collaborative, autonomous, and replacing. Emphasizes information sharing and the central role of trust in AI technology.
Paper III: Empirical study in a grocery retail supply chain showing that sharing downstream data with a neutral AI provider for joint forecasting reduces forecasting errors by ~40% and yields more stable forecasts.
Overall message: AI works best when it augments human decision-making and is paired with strong organizational capabilities and interfirm data-sharing arrangements.

Data & Methods

Paper I: Systematic literature review of management/supply-chain/AI studies; synthesis produced the Bullwhip-Smoothing-Framework (BSF) identifying managerial enablers.
Paper II: Conceptual development of an AI Collaboration Framework (five-stage taxonomy) informed by theory on human–AI interaction, information sharing, and trust.
Paper III: Empirical field study in a grocery retail supply chain implementing joint AI forecasting via a neutral AI provider. Partners shared downstream demand data to train joint models; outcomes measured by forecasting error reductions and stability of predictions (reported ≈40% error reduction).
Cross-study approach: Combination of literature synthesis, theoretical framework building, and an applied empirical intervention to test practical effects.
Limitations noted by the author: empirical evidence comes from a specific sector/context (grocery retail); details on model architectures and exact forecasting metrics are context-dependent and may require replication in other settings.

Implications for AI Economics

Value creation & cost savings: AI-enabled joint forecasting that reduces forecasting error and variability can lower inventory costs, reduce stockouts/overstock, and improve supply-chain resilience—improving firm profitability and consumer welfare.
Complementarity with human capital: The finding that AI is most effective when complementing rather than replacing human expertise implies complementarities between AI technology and managerial/labor skills; investments in digital skills and leadership matter for returns to AI.
Data as an economic input: Downstream demand data shared across partners becomes a critical input whose pooling increases model performance and reduces externalities caused by information fragmentation. This raises questions about incentives, pricing, and contract design for data sharing.
Role of intermediaries & market structure: Neutral AI providers that mediate joint forecasting can internalize coordination benefits and reduce distrust among firms, altering the market for forecasting services and potentially creating new platform intermediaries.
Trust, governance and policy: Trust in AI systems and governance structures to manage data privacy, intellectual property and antitrust concerns are central—policy interventions may be needed to facilitate secure data collaboration while preventing data-market abuses.
Research and measurement agenda: Economic research should quantify macro-level welfare impacts of reduced bullwhip (sector-wide inventory and service-cost savings), study incentive-compatible data-sharing mechanisms, evaluate generalizability across industries, and model labor-market effects given AI–human complementarities.

Assessment

Paper Typequasi_experimental Evidence Strengthmedium — The dissertation combines a systematic literature review and a conceptual framework with an empirical field study that reports a large (≈40%) reduction in forecasting errors — direct and relevant evidence that AI can reduce the bullwhip effect — but the empirical study is limited to a single grocery retail supply chain with no clear randomization, raising concerns about selection, confounding, and external validity. Methods Rigormedium — Strengths include a systematic review, clear conceptualization of AI–human roles, and an applied field evaluation using out-of-sample forecast comparisons; weaknesses include likely non-randomized treatment, limited transparency about sample size/time horizon/modeling details in the summary, and potential unobserved confounders and implementation heterogeneity that limit causal claims and replication. SampleThree-part cumulative dissertation: (1) systematic literature review of management/SCM studies on AI and bullwhip mitigation; (2) conceptual development of an AI Collaboration Framework; (3) an empirical field study in a grocery retail supply chain where downstream demand data were shared with a neutral AI provider to train joint forecasting models — the empirical sample consists of retailer sales/demand data (SKUs/stores/time series) and corresponding baseline forecasts, though exact counts of SKUs, stores, and time periods are not specified in the summary. Themesproductivity human_ai_collab org_design IdentificationQuasi-experimental field evaluation comparing out-of-sample forecast accuracy from a joint AI model trained on pooled downstream demand data (via a neutral AI provider) against firms' baseline forecasts; identification relies on holdout/test comparisons of forecast errors rather than random assignment. GeneralizabilitySingle industry (grocery retail) — may not generalize to manufacturing, high-tech, or B2B supply chains, Single supply chain / geographic context — country-level and market-structure differences could change effects, Study relies on a neutral external AI provider — results may differ when AI is developed/hosted in-house or by competitors, Firms willing to share downstream data may be atypical (selection bias), Unclear if results hold across different product types (seasonal, intermittent demand) or over longer horizons, Implementation- and model-specific effects (architecture, feature engineering) may limit replication

Claims (10)

Claim	Direction	Outcome	Confidence & Evidence	Details
The bullwhip effect, defined as the amplification of demand variability as orders move upstream, remains a persistent challenge in the field of supply chain management, causing inventory imbalances, service disruptions and financial losses. Organizational Efficiency	negative	inventory imbalances, service disruptions, financial losses	Reading fidelity high Study strength medium	not reported 0.48
Advances in AI have the potential to mitigate the bullwhip effect by integrating multiple data sources and supporting collaborative decision-making. Organizational Efficiency	positive	mitigation of bullwhip effect (reduced demand amplification)	Reading fidelity high Study strength speculative	not reported 0.08
The management literature has not yet clarified how AI can be effectively deployed to address the bullwhip effect. Adoption Rate	null_result	clarity/coverage of management literature on AI deployment for bullwhip mitigation	Reading fidelity high Study strength medium	not reported 0.48
Paper I proposes the Bullwhip-Smoothing-Framework (BSF), which highlights digital skills, leadership commitment and collaboration as essential management pillars for reducing the bullwhip effect. Organizational Efficiency	positive	importance of digital skills, leadership commitment and collaboration for bullwhip mitigation	Reading fidelity high Study strength medium	not reported 0.48
Paper II proposes an AI Collaboration Framework comprising five stages: complementary, augmentative, collaborative, autonomous and replacing. Task Allocation	neutral	classification of AI-human collaboration modes	Reading fidelity high Study strength speculative	not reported 0.08
Paper II finds that AI can support information sharing and trust among partners and stresses the need for trust in AI technology for successful deployment. Decision Quality	positive	information sharing and trust as facilitators of AI deployment	Reading fidelity high Study strength medium	not reported 0.48
Paper III presents an empirical study of joint AI forecasting in a grocery retail supply chain demonstrating that sharing downstream data with a neutral AI provider reduces forecasting errors by approximately forty percent and produces steadier predictions. Error Rate	positive	forecasting error (accuracy) and prediction steadiness	Reading fidelity high Study strength medium	approximately forty percent 0.48
Combining AI with organizational competencies (digital proficiency, leadership development and collaboration) can effectively reduce the bullwhip effect. Organizational Efficiency	positive	reduction in bullwhip effect (demand variability amplification)	Reading fidelity high Study strength medium	not reported 0.48
AI is most effective when it complements rather than replaces human expertise. Task Allocation	positive	effectiveness of AI deployment relative to complementarity vs. replacement	Reading fidelity high Study strength medium	not reported 0.48
AI-driven joint forecasting is most effective when partners share downstream demand data to train joint models. Error Rate	positive	forecast accuracy/error reduction through shared downstream demand data	Reading fidelity high Study strength medium	not reported 0.48