The Outlook of Artificial Intelligence in Manufacturing and Value Chains

Kiva Allgood¹, Devendra Jain¹ and Benedikt Gieger¹

1 Centre for Advanced Manufacturing & Supply Chains, World Economic Forum, Cologny/Geneva, Switzerland

E-mail: mailto:kiva.allgood@weforum.org

Status

AI has evolved from a promising technology to a transformative force, fundamentally reshaping global manufacturing and value chains. Over the past decade, AI and its applications have matured, driven by advances in data availability, algorithms, and compute power. As a result, manufacturers are increasingly recognizing AI’s potential to drive step-change improvements in efficiency, sustainability, and resilience when deployed at scale. The World Economic Forum’s Global Lighthouse Network¹, a community of advanced manufacturing sites, serves as a compelling showcase of such AI-enabled improvements.

The early use cases of AI focused on predictive maintenance and quality control. Today, AI’s integration spans the full value chain: demand sensing, supply planning, autonomous intralogistics, energy optimization, and dynamic scheduling. Notably, much of the current impact still stems from conventional AI models, which

continue to drive significant gains-often exceeding 50% in conversion cost, cycle times and defect rates². Importantly, AI is no longer a siloed technology; it is becoming embedded across fit-for-purpose intelligent systems that are digital, adaptive, or autonomous.

In parallel, the entire manufacturing industry faces an unprecedented confluence of pressures: labor shortages, climate challenges, geopolitical tensions, and sustainability imperatives³. AI has the potential to close productivity gaps and demographic challenges, localize production, and decarbonize industrial operations. Advances in AI-driven simulation, self-learning agents, and hybrid human-AI collaboration models promise to redefine how products are designed, made, and moved⁴.

Despite progress, the journey is far from complete. While some firms are moving toward full-scale deployment, many remain stuck in isolated pilots, hindered by fragmented data ecosystems, legacy infrastructure, talent shortages, or strategic misalignment. Bridging this gap will require scalable digital solutions, sustainability and resilience frameworks, strong data-management and a shift in workforce capabilities⁵.

Looking ahead, the focus must shift from experimentation to scaled impact. For manufacturers, that includes positioning themselves along a transformation continuum that reflects the evolving integration of AI into industrial systems. This journey typically unfolds in three progressive phases: digital, adaptive, and autonomous. In the digital phase, firms focus on building foundational capabilities such as connected data infrastructures, real-time visibility, and process automation. In the adaptive phase, AI is leveraged for scenario simulation, predictive insights, and dynamic decision-making, enabling responsiveness to changing market conditions. The autonomous phase marks the emergence of self-optimizing, self-healing operations, where for example AI agents manage complex networks with minimal human intervention⁶. As manufacturers navigate this continuum, those who successfully harness AI as a strategic enabler, will define the next era of intelligent and sustainable value creation.

Current and future challenges

Scaled AI adoption is impeded by a set of interrelated technological, organizational, and ethical barriers, also shown in figure 1. Working with global industry leaders and lighthouse factories, the World Economic Forum recognized a consistent set of hurdles that must be addressed to unlock AI’s next wave of transformative impact:

• Data & Digital Core: Despite the abundance of operational data, much of it remains siloed across departments, limiting end-to-end visibility. Many organizations operate heterogeneous IT systems (including legacy platforms, on-premise databases, and disparate cloud services) that were never designed for AI integration. The lack of interoperability and standardized data models inhibits development of scalable AI applications
• Governance, Ethics & Transparency: Accountability, fairness, and transparency is key when using AI, but the ‘black-box’ nature of many AI algorithms complicates efforts to understand, audit, or explain decisions. Biases embedded in training data or model design can result in discriminatory outcomes, potentially affecting suppliers, workers, or product quality. These risks are amplified by the rapid pace of AI innovation, with new models and capabilities emerging almost daily. Therefore, it is increasingly difficult for manufacturers to assess, validate, and govern these systems effectively⁷.
• Scaling beyond Pilots: Many firms struggle to translate proof-of-concept initiatives into enterprise-wide platforms due to a lack of clear return on investment, or integration issues with legacy systems. This

creates a paradox where firms acknowledge AI's strategic importance but underinvest in its full deployment. Repeated experimentation without systemic impact can also lead to a pilot fatigue, where stakeholders become disillusioned with AI’s promised benefits.

• Talent & Organizational Readiness: Scaling AI in manufacturing also demands a significant shift in workforce capabilities and organizational culture. The skills required to develop and operate AI systems
• ranging from data science and AI literacy to ethical reasoning and systems thinking needed for effective human-machine collaboration -are not yet widely distributed across the industrial workforce. Addressing this gap will require substantial change management efforts⁸.
• Transformation Complexity: Compounding these challenges is the growing complexity of strategic transformation itself. Manufacturers are increasingly expected to align their AI efforts with both sustainability and resilience objectives. This shift requires the simultaneous optimization of efficiency, environmental impact, and adaptability⁹. Trade-offs between goals like rapid delivery versus carbon reduction can be managed through advanced AI-driven optimization and decision support. However, most organizations lack the cross-functional structures needed to orchestrate such a triple transformation. Addressing these interdependent challenges is essential to unlocking the full potential of AI in manufacturing. Figure 1. Five principal challenges that organizations face in the context of digital transformation

and future readiness.

Advances in science and technology to meet challenges

In response to the multifaceted challenges facing AI adoption in manufacturing, advances in science and technology must be directed to solving persistent barriers. A new generation of scientific and technological advancements is emerging, pushing the boundaries of what’s possible in industrial settings:

• Domain-specific Industrial Foundation Models: Unlike large, general-purpose foundation models, smaller, domain-specific models trained on manufacturing-specific data such as machine logs and process parameters are on the rise. Their niche focus allows for more accurate, context-aware predictions while significantly reducing the computational resources and energy typically required by

large-scale models¹⁰. Their compact size enhances deployability at the edge like on shop floors or within connected machinery, where latency, bandwidth, and data privacy are critical concerns.

• Explainability Tools: The growing availability and integration of explainability tools allows interpretations of AI decisions. In industrial environments, where safety, compliance, and trust are paramount, the ability to understand why an AI system has made a specific recommendation is essential. Explainable AI (XAI) techniques enable users to trace outcomes back to input factors and assumptions. This transparency builds trust, facilitates regulatory compliance, and allows human experts to validate AI outputs when necessary, ensuring that automation enhances, rather than undermines, operational integrity.
• Agentic Systems: One of the most significant advancements is the development of intelligent operations through AI agents, both virtual and embodied. Virtual agents operate within software environments, while embodied agents perform increasingly sophisticated physical tasks on the factory floor¹¹. An illustrative example is an AI agent that autonomously manages shop floor disruptions and in cases like machine downtime, the AI agent proactively reschedules production and orchestrates material flow in real time. This system delivers contextual insights to supervisors in natural language and facilitates swift, informed responses, reinforcing trust between human and machine agents.
• Human-AI Collaboration Systems: As AI systems take over routine, deterministic tasks, human roles are shifting toward oversight, exception management, and creative problem-solving. The relationship is increasingly symbiotic: intelligent systems handle complexity and scale, while humans provide contextual judgment, ethical evaluation, and adaptability in unforeseen situations. Also, the interaction between human and AI is becoming a more effective collaboration. XAI, natural language interfaces, and augmented reality tools allow frontline workers to interact with AI systems intuitively. This fosters trust and bridge the digital skill gap by embedding AI into existing workflows¹². Together, these advances are not only addressing the current limitations of AI adoption but are also laying the groundwork for a new era of intelligent, adaptive, and resilient manufacturing.

Concluding remarks

AI is poised to redefine manufacturing and value chains, emerging as the fundamental operating system of the next industrial era. It is essential for manufacturers to scale beyond pilots and build a strong digital foundation purpose-built for AI, reducing integration efforts.

Success will also hinge on aligning AI deployment with broader transformation goals and keeping humans at the core of this transformation. The convergence of digitalization, sustainability, and resilience has led to the emergence of a new model for industrial transformation – one that is enabled and orchestrated by AI. Rather than treating them as three separate domains, manufacturers are unifying them into one transformation, a convergence that can be seen as a “triple transformation”. This system effectively creates “self-healing” operations that can anticipate and mitigate shocks before they cascade through the value network¹³. In an era defined by complexity, volatility, and systemic constraints, manufacturers who embrace AI as the enabler of their transformation will be the ones to lead.

Acknowledgements

Disclaimer: the original content of the article (and not any third-party content referenced or included) is covered by the CC BY 4.0 licence. For avoidance of doubt, the views expressed are solely those of the authors and do not represent the official position of the authors’ employers.

References

[1] World Economic Forum. (2025). World Economic Forum: Global Lighthouse Network. [2] World Economic Forum. (2025). WEF Global Lighthouse Network: The Mindset Shifts Driving Digital Impact and Scale in Digital Transformation. [3] World Economic Forum. (2023). World Economic Forum: A Global Rewiring: Redefining Global Value Chains for the Future. [4] World Economic Forum. (2025). Navigating the AI Frontier: A Primer on the Evolution and Impact of AI Agents. [5] World Economic Forum. (2025). AI in Action: Beyond Experimentation to Transform Industry. Geneva/Cologny: World Economic Forum. [6] Metzger, D. (2024, 08 22). SAP. Retrieved from Modernizing Supply Chains: The Autonomous AI-Driven Future: https://news.sap.com/2024/08/modern-autonomous-ai-supply-chain/ [7] World Economic Forum. (2025). AI in Action: Beyond Experimentation to Transform Industry. Geneva/Cologny: World Economic Forum. [8] World Economic Forum. (2025). Putting Talent at the Centre: An Evolving Imperative for Manufacturing. Geneva/Cologny: World Economic Forum. [9] Christmann, A.-S., Crome, C., Graf-Drasch, V., Oberländer, A., & Schmidt, L. (2024, 01 23). The Twin Transformation Butterfly. Retrieved from https://www.researchgate.net/publication/377631322_The_Twin_Transformation_Butterfly [10] Zaika, K. (2024, 09 18). Forbes. Retrieved from Why Choose A Domain-Specific LLM For Your Business?: https://www.forbes.com/councils/forbestechcouncil/2024/09/18/why-choose-a-domain-specific-llm-for-your-business/ [11] World Economic Forum. (2025). Frontier Technologies in Industrial Operations: The Rise of AI Agents. Geneva/Cologny: World Economic Forum. [12] World Economic Forum. (2025). Frontier Technologies in Industrial Operations: The Rise of AI Agents. Geneva/Cologny: World Economic Forum. [13] Marco, N. d. (2023, 09 29). Forbes. Retrieved from Building Resilient Organizations With AI: https://www.forbes.com/councils/forbestechcouncil/2023/09/29/building-resilient-organizations-with-ai/