What’s Really Holding Back AI Adoption in Manufacturing and Supply Chains? (Part 2 — Technical & Data Challenges)

This article is part of a 3-part series exploring the key barriers preventing successful AI adoption in manufacturing and supply chains, based on recent research and industry insights. In Part 1, I examined organizational and cultural barriers. This second article focuses on the technical and data-related challenges that often prevent AI solutions from scaling effectively.

Without high-quality, integrated, and accessible data and without the infrastructure to support real-time analytics, even the most sophisticated AI models will struggle to deliver business value. Data is the fuel of AI. Yet in many manufacturing and supply chain environments, data landscapes are fragmented, legacy systems dominate operations, and information silos persist across functions.

In this second article, I turn to these technical and data-related challenges and unpack why they remain such a persistent obstacle, and how companies can start building the foundations needed for successful, scalable AI adoption.

Technical Hurdles in AI Integration

While AI holds enormous promise for improving decision-making in manufacturing and supply chains, its success depends on solid technical foundations. Many organizations underestimate the complexity of integrating AI into existing operational environments, especially in industries with long-standing legacy systems, fragmented data landscapes, and rigid workflows.

The literature consistently highlights four critical technical barriers that prevent AI projects from moving beyond isolated pilots (Kar et al., 2022; Kaggwa et al., 2024).

1. Legacy Systems and Infrastructure: One of the most common challenges is the reliance on outdated legacy systems that were never designed to support modern AI applications. These systems often lack the flexibility, connectivity, and processing power required to feed real-time data into AI models. For example, a manufacturing company might operate with decades-old ERP systems that store production data in non-standard formats or siloed databases. Integrating an AI-driven predictive maintenance tool into this environment may require expensive middleware solutions or even a complete infrastructure overhaul which could add time, cost, and complexity to AI projects.

2. Data Silos and Inconsistencies: Data is essential for AI, yet in many supply chain environments, data is dispersed across functions, locations, and legacy IT systems. Different departments collect and store data in different formats, leading to inconsistencies that hinder AI’s ability to generate accurate insights. For instance, the procurement team might manage supplier data in one system, while logistics uses another, with no shared data standards. When an AI tool is introduced to optimize inventory or forecast demand, these inconsistencies create errors, misalignment, and distrust in the system’s recommendations.

3. Scalability Issues: Initial AI pilots often show promising results in controlled environments. However, scaling AI solutions across complex supply chain operations presents new difficulties. The computational requirements grow, the volume and variability of data increase, and ensuring consistent performance across multiple sites becomes challenging. Consider a company that successfully uses AI for warehouse optimization in one location. Scaling that solution globally may require addressing very different infrastructure conditions, data availability, or operational constraints, all of which require additional development, resources, and local customization.

4. Integration with Existing Workflows: Finally, even technically sound AI solutions can face resistance if they disrupt established operational workflows. Employees may perceive new tools as cumbersome or intrusive, especially if the AI system changes familiar processes without demonstrating clear added value. For example, a supply chain planning team accustomed to working with spreadsheets might resist adopting an AI-based forecasting dashboard that requires a new interface, new routines, and new decision protocols, unless the benefits are made explicit and the transition is well managed.

Data Quality and Management Challenges

Even with the right infrastructure in place, AI adoption in manufacturing and supply chains cannot succeed without addressing fundamental data quality and management issues. AI systems are only as good as the data they rely on, and many organizations struggle to provide the clean, consistent, and timely data that AI needs to generate reliable insights.

Research by Kar et al., (2022), Kaggwa et al. (2024) and Stone et al. (2020) highlights four key data-related barriers that frequently undermine AI initiatives.

1. Data Accuracy and Completeness: AI models require high-quality, accurate, and complete datasets to function effectively. If the data feeding into an AI system is outdated, incomplete, or incorrect, the model’s outputs, no matter how sophisticated, will be flawed. For example, a manufacturer using AI to predict equipment failures needs reliable sensor data on machine performance. If some sensors are faulty or certain operational data is missing, the AI may produce inaccurate predictions, leading either to unnecessary maintenance costs or unexpected breakdowns, both of which erode trust in the system.

2. Real-Time Data Processing: In dynamic supply chain environments, static or historical data is often insufficient. AI systems need access to real-time data streams to provide timely and actionable insights, such as detecting supply disruptions or forecasting demand spikes. For example, a manufacturer may implement an AI-driven production planning system designed to optimize scheduling based on real-time machine availability and order changes. However, if machine performance data or order updates are only synchronized once per shift or once per day, the AI system operates with outdated information, reducing its ability to respond quickly to unforeseen disruptions on the shop floor.

3. Data Security and Privacy: The increasing use of AI in operational decision-making also raises significant concerns about data security and privacy. Supply chains often involve sensitive data on suppliers, customers, pricing, and operational performance consisting of data that must be protected, especially when shared across borders or with third parties. For example, a logistics company using AI for real-time route optimization across multiple countries must navigate different data protection regulations, such as GDPR in Europe. This means ensuring that personal data, such as driver locations or customer delivery addresses, is collected, processed, and stored securely in compliance with local laws. In some cases, the company may need to anonymize data, store it within specific geographic locations, or obtain explicit user consent. These requirements create significant complexity because AI systems often rely on centralized data processing and real-time analytics, while privacy regulations may require data to remain within specific jurisdictions, limit how long it can be stored, or enforce strict controls on access and usage.

4. Data Governance: Underlying all these challenges is the need for strong data governance, such as clear policies, processes, and accountability for how data is collected, managed, and used across the organization. Without defined data ownership and stewardship roles (responsible for the quality, consistency, and proper use of data), AI initiatives risk being undermined by inconsistent data practices or unresolved disputes over data access. For instance, if operational data is owned by the production team but needed by the AI development team, unclear governance structures can delay projects or lead to fragmented data usage. Strong data governance ensures that data integrity is maintained throughout its lifecycle, from collection to analysis, and that all stakeholders understand their responsibilities in managing data assets.

Strategies to Overcome Technical and Data Challenges

Overcoming technical and data-related barriers to AI adoption in manufacturing and supply chains requires deliberate action in several key areas. The literature emphasizes that successful AI integration depends on building strong data foundations, aligning technology with operational needs, and investing in people and processes to support long-term adoption (Kar et al., 2022; Kaggwa et al., 2024; Stone et al., 2020).

• Modernize IT Infrastructure: AI systems require flexible, scalable infrastructure capable of handling large volumes of real-time data. Many companies are moving toward cloud-based platforms that support seamless integration of AI tools and enable faster processing across distributed operations, helping to reduce the burden on local IT resources, improve data accessibility, and enhance system resilience.

• Implement Data Integration Solutions: To overcome data silos and inconsistencies, organizations need data integration tools that consolidate information from different systems into a unified, accessible platform. Integrating procurement, production, and logistics data provides the comprehensive, high-quality datasets that AI systems require to detect patterns, optimize operations, and generate accurate predictions across the supply chain.

• Develop Robust Data Governance Frameworks: Strong data governance ensures that data is reliable, consistent, and compliant with regulatory requirements. Defining clear policies on data standards, ownership, and quality controls supports operational decision-making and compliance efforts. Assigning data stewardship roles within operational teams helps maintain data integrity and fosters accountability in managing data assets.

• Pilot AI Initiatives: Starting with small-scale AI pilots allows organizations to test applications in a controlled setting and refine their approach before scaling. Piloting an AI-powered route optimization tool in one distribution center, for instance, helps identify integration challenges, build user confidence, and generate early learnings to inform broader implementation.

• Foster Cross-Functional Collaboration: AI initiatives benefit from close collaboration between technical teams and operational staff. Involving supply chain planners, warehouse managers, and data engineers in the design of AI tools ensures that the output aligns with operational realities, improves usability, and facilitates smoother adoption across teams.

• Invest in Training and Change Management: Equipping employees with the skills to understand and work with AI systems is essential for long-term adoption. Workforce development initiatives focused on data literacy and change management help operational teams interpret AI-generated insights, integrate new tools into their daily routines, and foster a culture of experimentation and innovation .

By proactively addressing these technical and data challenges, manufacturing and supply chain organizations can harness the full potential of AI, leading to enhanced decision-making capabilities, increased efficiency, and a competitive edge in the market.

Technical and data challenges may not be the most exciting part of AI adoption, but they are often the reason projects get stuck or fail to scale. No amount of ambition or vision can overcome poor infrastructure, scattered data, or unclear governance. Getting these basics right is what turns promising pilots into real business value.

In the final article of this series, I will turn to the ethical and governance challenges that organizations need to address as AI becomes more embedded in strategic decision-making. Building trust and ensuring responsible AI adoption will be just as important as the technology itself.

References

Kar, S., Kar, A. K., & Gupta, M. P. (2022). Modeling Drivers and Barriers of Artificial Intelligence Adoption: Insights from a Strategic Management Perspective. Intelligent Systems in Accounting, Finance and Management, 28(4), 217–238. https://doi.org/10.1002/isaf.1503

Kaggwa, S., Eleogu, T. F., Okonkwo, F., Farayola, O. A., Uwaoma, P. U., & Akinoso, A. (2024). AI in Decision Making: Transforming Business Strategies. International Journal of Research and Scientific Innovation, 10(12), 423–432. https://doi.org/10.51244/IJRSI.2023.1012032

Stone, M., Woodcock, N., Machtynger, J., & Heppel, M. (2020). Artificial Intelligence in Strategic Marketing Decision Making: A Review. The Bottom Line, 33(3), 205–214. https://doi.org/10.1108/BL-03-2020-0022