Designing the Future of Intelligent Manufacturing with Generative AI

As Industry 4.0 progresses, the essential components for AI integration are being established, including data and technology infrastructure, a skilled workforce, and operational models. Industrial AI has reached an unprecedented level of maturity, supported by strong data and technology infrastructure. This progress is paving the way for the practical application of intelligent and autonomous robots and systems.

The WEF/McKinsey technology pyramid (technology stack) for next-generation intelligent manufacturing includes the following layers (see Figure 1):

(1) Foundational Data, Connectivity, and Computing Tools
Examples: Cloud computing, edge computing, 5G/6G communications, data lakes
(2) System-Level Digitization of Planning and Control
Examples: Manufacturing Execution Systems (MES), Customer Relationship Management (CRM), Product Lifecycle Management (PLM)
(3) Process Automation and Production Process Innovation Tools
Examples: Collaborative robots (cobots), flexible robots, AGVs, drones, 3D printers
(4) Operator or Process-Level Digital Worker Productivity Tools
Examples: AR/VR, wearables, exoskeletons, dashboards
(5) Machine Intelligence Technologies for Predicting, Optimizing, and Enhancing Decision-Making
Examples: heuristic models, applied AI, generative AI

Intelligent manufacturing concepts and technologies are now widely adopted, with Industry 4.0 initiatives advancing across various sectors. The WEF and McKinsey's Global Lighthouse Network serves as a model for future development, with AI playing an increasingly important role(*2).

These Lighthouses are leading the large-scale application of next-generation intelligent manufacturing technologies, driving improvements in finance, operations, and sustainability. 172 sites were certified by the end of 2024. Over the past six years, the Network has demonstrated over 1,000 use cases, with 139 adopted by 132 companies as of January 2023. Notably, AI-related use cases within the top five WEF-certified categories grew from less than 20% in 2018 to nearly 60% by 2023. Implementing these AI use cases has yielded significant returns: 2-3x productivity increases, 50% better service levels, 99% defect reduction, and 30% lower energy consumption.

The WEF Lighthouse assessment quantitatively evaluates technical potential, business performance impact, and sustainability. The framework comprises five major categories and ten or more sub-categories. While impact varies across lighthouses, proper implementation consistently delivers expected results, motivating other companies to adopt intelligent manufacturing. Consequently, 64% of lighthouses demonstrate positive environmental impact by integrating multiple use cases alongside improvements in demand chain agility, customer centricity, supply chain resilience, and productivity/speed. Sustainable Lighthouses prove that digital technology can drive both sustainability and competitiveness.

Global Lighthouse use cases, KPIs, and operational mechanisms provide valuable benchmarks, serving as a model for implementing both conventional AI and more recent generative AI, particularly in software infrastructure development.

Industry 4.0 initially prioritized improving manufacturing efficiency. This focus is now evolving towards customer value. Next-generation intelligent manufacturing must encompass the entire enterprise, from supply chain and manufacturing to R&D, marketing, sales, and customer service.

Meanwhile, conventional AI, while strong in predictability and consistency, faces limitations with unstructured, real-time data. Generative AI, with its adaptability, flexibility, and creativity, offers a powerful solution for areas like R&D, marketing, sales, customer service, data infrastructure, and HR.

Three case studies illustrate generative AI's potential(*3) :

GE Appliances: Moving beyond smart appliances, GE is integrating generative AI into its SmartHQ app. Features like Flavorly analyze customer ingredients to generate recipes, simplifying cooking, saving grocery costs, and reducing food waste. This exemplifies embedding generative AI directly into products.

Bosch: This automotive supplier used generative AI to create synthetic training data for automated optical inspection of electric motor stators. Generating over 100x more images than available real data, this reduced project time from years to six months, improved inspection quality, and yielded six-figure annual productivity gains, showcasing the synergy between generative and conventional AI.

ACG Capsules: To address evolving workforce skills, ACG deployed a custom generative AI assistant in just two weeks, adding training and gamification. Within five weeks, nearly 75% of operators and technicians used the assistant, reducing mean time to repair (MTTR) by 30-40% and earning them Global Lighthouse recognition.

Many Global Lighthouse examples utilize conventional AI for specific tasks at the process level, often requiring separate models for each use case, including some "lights out" operations. However, conventional AI's rule-based nature limits its adaptability and flexibility.

Generative AI, in contrast, offers greater adaptability, flexibility, and creativity, driving improvements in productivity, personalized customer experiences, and human imagination. Current applications, however, are primarily limited to simple question-and-answer interactions.

Looking ahead, generative AI is evolving into AI agents capable of understanding context, planning workflows, connecting to external resources, and taking action to achieve goals. This includes collaboration, process orchestration, and autonomous decision-making. Multi-agent systems comprising AI agents, task-specific agents, and orchestration agents are under active development.

This vision is already being realized. Fujitsu's "Kozuchi AI Agent," for example, participates in meetings and negotiations, providing information and suggesting actions. It can analyze statements like "Asian sales are half of last year's," displaying detailed sales data and facilitating productive discussions. Fujitsu has also developed an AI agent for video analysis to support safe and efficient work sites (warehouses, factories, etc. See Figure 2)(*4) . In the future, Fujitsu plans to expand its AI agent offerings to include specialized agents for production management and other tasks.