From Low-Budget Pilots to Scalable AI in Operations and Supply Chain Management: a Roadmap Built on the “Blueprints” of Implementation

Purpose of the paper

The dominant narrative around artificial intelligence (AI) in operations and supply chain management (OSCM) overemphasises large-scale, capital-intensive digital transformations (Choi et al., 2018; Belhadi et al., 2024). However, most organisations, both small- and medium-sized enterprises (SMEs) and large enterprises (LEs), begin their AI journeys with deliberately modest, low-budget pilots that allow them to learn, test feasibility, and preserve option value before committing to larger investments (Moeuf et al., 2018; Wu and Pagell, 2011). The purpose of this paper is to explore why some of these frugal yet pragmatic pilots successfully scale into reliable, production-grade systems while others stall. We argue that scaling success depends on systematically upgrading the operational foundations, termed the “blueprints” of implementation (data infrastructure, hardware, integration, and governance), rather than relying solely on algorithmic sophistication (Sculley et al., 2015).

Design/methodology/approach

Drawing on a portfolio of 15 industrial AI projects as illustrative field evidence, we analyse implementations spanning image recognition, data mining, and agentic AI. The projects were conducted across both SMEs (7 projects) and LEs (8 projects). All 15 projects began as low-budget pilots and achieved technical feasibility, but their ultimate scaling outcomes varied. By examining these divergent paths, we identify recurring bottlenecks and patterns, distil them into four distinct implementation archetypes, and map them to a proposed five-stage implementation roadmap.

Findings

Out of the 15 projects, 10 (67%) successfully scaled to production. Four key patterns emerged that support the logic of our roadmap. First, data readiness is the single strongest predictor of scaling success: projects starting with structured data scaled 100% of the time, while those with messy or partial data scaled in only 25% of cases. Second, operational complexity is the true barrier, not algorithmic complexity. Counterintuitively, high-complexity projects reached production more frequently (83%) than low-complexity ones (33%), because complex algorithms required deliberate investment in operational blueprints, whereas simple algorithms often led to complacency. Third, there is a scaling divergence by enterprise size: LEs scaled at a higher rate (75%) than SMEs (57%), which is tied to prior maturity in IT infrastructure and governance rather than just budget differences (Moeuf et al., 2018; Tortorella and Fettermann, 2018). Finally, bottlenecks consistently concentrate in the blueprints (integration and data governance accounted for the majority of scaling hurdles), and never in the AI model itself.

These findings are categorised into four "blueprints" archetypes:

• Archetype A: Data Foundation. Developing models (e.g., clustering for lead times) stalls for firms with incomplete records, whereas firms with high-quality, structured data successfully integrate the same algorithm into operations (Wamba et al., 2015).

• Archetype B: Hardware and Performance. Consumer-grade pilots (e.g., Raspberry Pi object detection) prove valuable but require targeted upgrades to industrial hardware to meet real-time production constraints (Dubey et al., 2020).

• Archetype C: Integration and Workflow. Cloud-based NLP models can easily handle discrete tasks, but scaling requires extensive orchestration layers to handle heterogeneous document pipelines and privacy constraints (Holmström and Partanen al., 2014; Sculley et al., 2015).

• Archetype D: Governance and AI Guardrails. Agentic AI carries hallucination risks (Baryannis et al., 2019), meaning scaling requires strict validation mechanisms before outputs enter enterprise systems (Davis et al., 2024).

Value/Originality

This paper challenges the "glamour" narrative of AI in OSCM by providing a pragmatic, operations-centric framework. It introduces a novel five-stage stage-gate roadmap: (1) Data Foundation and Governance, (2) Feasibility Pilot, (3) Value-Validated Pilot, (4) Industrialization and Scaling Preparation, and (5) Production Deployment and Replication. It establishes that while early stages (1–3) suit low-budget approaches to validate value with minimal capital exposure (Ferreira et al., 2016), the critical transition occurs at Stage 4, where upgrading the operational blueprints (hardware, IT/OT integration, data pipelines, and MLOps) is mandatory (Ivanov et al., 2019). The paper shifts the focus from pure algorithmic innovation toward the operational conditions that enable real-world implementation.

Research limitations/implications

The portfolio approach serves as illustrative field evidence rather than formal hypothesis testing. The findings open a research agenda for OSCM scholars. Future research should formalise the economics of low-budget pilots as real options versus full-scale deployment (Wu and Pagell, 2011). Scholars should also investigate differences in organisational absorptive capacity between SMEs and LEs, develop agile governance frameworks for agentic AI that ensure reliability without compromising agility (Davis et al., 2024), and study data readiness as a managed operational resource subject to accumulation and degradation.

Practical implications

The five-stage roadmap and four archetypes function as a diagnostic instrument for practising managers, helping them locate a project on its implementation trajectory and identify exactly what is required to move it forward. Key practical rules include: do not attempt Stage 2 modelling without a credible Stage 1 data foundation; treat Stage 4 industrialization investments as real options to be exercised only after value is validated in Stages 2–3; map the workflow integration surface early, regardless of enterprise size; and, for agentic AI, establish human-in-the-loop fallback mechanisms and guardrails during Stage 4 before granting autonomous power.

References

Baryannis, G., Validi, S., Dani, S. and Antoniou, G., 2019. Supply chain risk management and artificial intelligence: state of the art and future research directions. International journal of production research, 57(7), pp.2179-2202.

Belhadi, A., Mani, V., Kamble, S.S., Khan, S.A.R. and Verma, S., 2024. Artificial intelligence-driven innovation for enhancing supply chain resilience and performance under the effect of supply chain dynamism: an empirical investigation. Annals of operations research, 333(2), pp.627-652.

Choi, T.M., Wallace, S.W. and Wang, Y., 2018. Big data analytics in operations management. Production and operations management, 27(10), pp.1868-1883.

Davis, A.M., Mankad, S., Corbett, C.J. and Katok, E., 2024. OM Forum—The best of both worlds: Machine learning and behavioral science in operations management. Manufacturing & Service Operations Management, 26(5), pp.1605-1621.

Dubey, R., Gunasekaran, A., Childe, S.J., Bryde, D.J., Giannakis, M., Foropon, C., Roubaud, D. and Hazen, B.T., 2020. Big data analytics and artificial intelligence pathway to operational performance under the effects of entrepreneurial orientation and environmental dynamism: A study of manufacturing organisations. International journal of production economics, 226, p.107599.

Ferreira, K.J., Lee, B.H.A. and Simchi-Levi, D., 2016. Analytics for an online retailer: Demand forecasting and price optimization. Manufacturing & service operations management, 18(1), pp.69-88.

Holmström, J. and Partanen, J., 2014. Digital manufacturing-driven transformations of service supply chains for complex products. Supply Chain Management: An International Journal, 19(4), pp.421-430.

Ivanov, D., Dolgui, A. and Sokolov, B., 2019. The impact of digital technology and Industry 4.0 on the ripple effect and supply chain risk analytics. International journal of production research, 57(3), pp.829-846.

Moeuf, A., Pellerin, R., Lamouri, S., Tamayo-Giraldo, S. and Barbaray, R., 2018. The industrial management of SMEs in the era of Industry 4.0. International journal of production research, 56(3), pp.1118-1136.

Sculley, D., Holt, G., Golovin, D., Davydov, E., Phillips, T., Ebner, D., Chaudhary, V., Young, M., Crespo, J.F. and Dennison, D., 2015. Hidden technical debt in machine learning systems. Advances in neural information processing systems, 28.

Tortorella, G.L. and Fettermann, D., 2018. Implementation of Industry 4.0 and lean production in Brazilian manufacturing companies. International journal of production research, 56(8), pp.2975-2987.

Wamba, S.F., Akter, S., Edwards, A., Chopin, G. and Gnanzou, D., 2015. How ‘big data’can make big impact: Findings from a systematic review and a longitudinal case study. International journal of production economics, 165, pp.234-246.

Wu, Z. and Pagell, M., 2011. Balancing priorities: Decision-making in sustainable supply chain management. Journal of operations management, 29(6), pp.577-590.

Purpose of this paper

Adoption of Artificial Intelligence (AI) as a tool to support various supply chain functions, processes and activities are surging in all industrial sectors including the Hospitality industry. This research assesses the role of Procure to Pay (P2P) Coupa integrated AI systems, its associated Digital Procurement Practices and their influence on the procurement cost and supply chain performance.

Design/methodology/approach

Based on theorization it is conceptualised that Procurement Efficiency is acting as a mediating relationship between the Independent and Dependent variable and their associations. An empirical survey questionnaire has been designed among the conceptualised constructs, and the survey was circulated specifically Hospitality Industry representatives in the United Arab Emirates which resulted in achieving 130 samples. The survey results were statistically analysed by using the Structural Equation Model using Smart PLS V4 software.

Findings

The Structural path model testing reveals that AI deployment via Coupa's P2P system (p < 0.001) and digital procurement practices (p < 0.001) positively influence operational efficiencies, which in turn significantly affect cost savings (p < 0.001). This supports the hypotheses that improved operational efficiencies lead to better cost outcomes, with operational efficiencies acting as a critical pathway through which AI and digital practices drive cost savings.

Value

Positive impact of AI Deployment and Digital Procurement Practices on Operational Efficiency and Cost Savings, reinforcing their critical role in driving operational and financial success in the hospitality sector. However, the limited influence of Operational Efficiency on Supplier Performance highlights a gap that requires further investigation, such as addressing industry-specific challenges like seasonal demand fluctuations and perishable inventory management. UAE hospitality industry being vulnerable to the global demand uncertainty (Elshaer et al. 2025), this research findings add value to the industry stakeholders.

Practical implications

The practical implications of this model emphasize the necessity for hospitality organizations to invest in employee training and seamlessly integrate AI technologies with existing systems to maximize benefits. While this research findings are inline with the Boston Consulting Group (BCG, 2026) study findings about AI influence in Hospitality industry inventory replenishment and customer experiences, there is considerable research gap still existing that warrant deeper investigation of several facets of its adoption level not just patchwork in various procurement stages.