Industrial AI is entering a new phase. While manufacturers have spent the past decade deploying machine learning systems to analyse production data and automate specific tasks, a new concept is beginning to reshape how factories operate: Agentic AI.
Bosch Connected Industry is among the companies exploring how intelligent software agents can work alongside human experts to coordinate complex manufacturing processes. Rather than focusing on isolated AI use cases, this approach aims to orchestrate entire production environments using interconnected digital agents.
In this exclusive interview with Norbert Jung, CEO of Bosch Connected Industry, he explains how Agentic AI differs from traditional industrial AI systems, where it is already delivering measurable results inside Bosch plants, and why semantic data structures and multi-agent systems could play a key role in the future of intelligent manufacturing.
Bosch Connected Industry is exploring how Agentic AI in manufacturing can coordinate machines, software platforms, and production systems through intelligent software agents.
MTN: Bosch has begun speaking about “Agentic AI” in manufacturing. In your view, what fundamentally differentiates Agentic AI from the industrial AI systems factories have been deploying over the past five years?
Norbert Jung:
The shift is fundamental. For the last five years, we’ve been using AI as a set of specific tools to automate narrow, repetitive tasks. Agentic AI is entirely different. We are no longer just automating tasks; we are orchestrating entire processes. These agents act as digital collaborators that can observe, reason, and proactively recommend actions to optimize the whole value chain. It’s the difference between a tool that does one job and an expert partner that helps you run the entire operation more intelligently.
MTN: Traditional AI systems generate insights. Agentic AI suggests autonomous decision-making and coordinated action. How far are we from factories where AI agents actively manage production processes rather than simply supporting human operators?
Norbert Jung:
Let me be very clear: a “lights-out” factory, run entirely by AI, is not our goal. Our strategy is built on what we call Manufacturing Co-Intelligence®, which is a deliberate choice to keep our human experts at the heart of our operations. The role of AI is to augment and amplify the capabilities of our people, not to replace them. The agents provide world-class analysis and powerful recommendations, but the final decision and ultimate responsibility will always rest with our skilled workforce. We see the future as a powerful synergy between human experience and machine intelligence.
MTN: Where are you already seeing measurable business impact from Agentic AI within Bosch’s own production network, whether in productivity, downtime reduction, quality improvement, or planning stability?
Norbert Jung:
The impact isn’t a future promise; it’s delivering real value today in our plants across Germany, Hungary, and India. For example, our Shopfloor AI Agent is slashing production downtime by helping our teams resolve disruptions three to five times faster. This translates to annual savings of approximately €850,000 per plant, from reduced machine down time.
MTN: You emphasize the importance of semantic data structures for Agentic AI. Why is a semantic layer essential, and what happens when companies attempt to deploy advanced AI agents without that foundation?
Norbert Jung:
A semantic layer is the non-negotiable foundation for any serious industrial AI strategy. It solves what we call the “Data Growth Paradox,” where having more data doesn’t create more value because it’s locked in silos. The semantic layer provides a common language, a single source of truth, that allows our agents to understand the context and relationships between data from hundreds of different systems. Without it, companies are building on sand. They get trapped in a “permanent construction site,” trying to connect data for every single new use case. It’s inefficient, it doesn’t scale, and the AI simply cannot deliver reliable, high-quality results.
MTN: Bosch has discussed multi-agent systems working together in manufacturing. Can you describe how these agents interact in a real production scenario, and what level of autonomy they currently operate with?
Norbert Jung:
Imagine a machine suddenly stops. In the past, this would trigger a lengthy, manual investigation. Today, our Shopfloor AI Agent is immediately notified. It autonomously analyzes historical data from a dozen systems to diagnose the root cause in seconds and provides the operator with clear, step-by-step instructions to fix it. Once resolved, it can automatically update the shift log and even trigger another agent to schedule a follow-up maintenance task. They operate with semi-autonomy within a strict “human-in-the-loop” framework. The agents can propose and prepare actions, but a human expert always makes the final call.
MTN: There is growing debate about AI autonomy in industrial settings. How do you define the right balance between agent-driven decisions and human oversight in safety-critical production environments?
Norbert Jung:
The balance is crystal clear: the human is, and will remain, the ultimate authority. In our model, the AI agent is an expert assistant that provides data-driven recommendations of the highest quality. But the human operator, with their years of experience, validates that advice and makes the final decision. Responsibility is never delegated to the machine. For safety-critical tasks where there is zero margin for error, we rely on deterministic systems, not probabilistic AI. This combination of intelligent guidance and non-negotiable human oversight is what makes our approach truly “industrial grade.”
MTN: Do you see Agentic AI becoming a competitive differentiator between European manufacturers, particularly as Manufacturing-X and sovereign data ecosystems evolve?
Norbert Jung:
Absolutely. Agentic AI is not just another technology; it is a strategic capability that will define competitiveness in manufacturing for the next decade. Sovereign data ecosystems like Manufacturing-X are critical because they allow us to train our agents on deep, proprietary production knowledge that isn’t publicly available. This creates a powerful competitive moat. By embedding our unique expertise into these agents, we can achieve levels of efficiency, agility, and resilience that our competitors simply cannot match.
MTN: Many AI initiatives remain contained within pilots. What needs to change organizationally and technically for Agentic AI to scale across multiple plants and international production networks?
Norbert Jung:
Scaling Agentic AI requires a dual transformation. Technically, you must commit to building a scalable semantic data foundation. This is the central nervous system that allows you to reuse data models and digital twins across your entire global network. You cannot scale without it. Organizationally, the change is even more profound. Leadership must move beyond isolated pilot projects and drive a strategic, top-down vision for an AI-powered enterprise. This also means choosing partnership over a “do-it-yourself” approach. By leveraging established frameworks, our teams can focus on creating business value, not on reinventing the complex underlying technology.
MTN: Looking ahead five years, how will Agentic AI reshape the role of production managers, maintenance engineers, and industrial engineers?
Norbert Jung:
The roles of our industrial experts will be elevated. Instead of spending their days firefighting and on routine monitoring, their responsibilities will shift toward supervising and collaborating with digital agents. They will become strategists, using AI-generated insights to focus on creative problem-solving and long-term process improvements. They will have the power to make data-driven decisions themselves, without needing a team of data scientists. Their deep domain expertise will be more valuable than ever as they take on the critical role of training, validating, and improving our AI systems.
MTN: As you look toward next year and beyond, what developments in Agentic AI should manufacturers be preparing for now?
Norbert Jung:
To stay ahead, manufacturers must act now. First, move beyond pilots. The time for small showcases is over; it’s time for strategic, enterprise-wide rollouts that are directly tied to improving business KPIs. Second, build your data foundation. A clean, structured, and semantically rich data layer is the single most important prerequisite for success. Finally, embrace partnership. The pace of innovation is too fast to go it alone. By collaborating with technology leaders, you can accelerate your transformation and focus your resources where they matter most: on building a more competitive and resilient manufacturing operation.
MTN Analysis
The rise of Agentic AI in manufacturing represents a shift from isolated AI tools toward intelligent digital agents that collaborate across production environments.
Bosch’s push toward Agentic AI highlights a shift taking place across modern manufacturing. Traditional industrial AI systems have focused on analysing machine data and delivering insights to engineers. Agentic AI expands this capability by introducing intelligent software agents that can coordinate actions across multiple systems within a factory.
Instead of monitoring a single machine or dataset, these agents analyse events across production environments and recommend coordinated responses. Bosch’s Shopfloor AI Agent resolving disruptions three to five times faster demonstrates how AI-driven operational intelligence can directly reduce machine downtime.
Reducing production interruptions remains one of the most valuable improvements manufacturers can achieve. Even small gains in uptime translate into significant financial impact. Bosch estimates savings of around €850,000 per plant annually when downtime is reduced through faster root cause analysis.
Another important theme is the role of data architecture. Many factories operate fragmented systems that store operational data in separate silos. Without a unified data structure, AI systems struggle to interpret relationships between machines, processes, and production events.
Another emerging trend is physical AI in robotics and automation, where machines use real-time data and AI models to adapt their movements and processes inside manufacturing environments.
Semantic data layers address this challenge by creating a common language across industrial systems. This enables AI agents to understand production context and support more complex decision making across the factory.
For many manufacturers, Agentic AI in manufacturing could become the next major step after predictive maintenance and machine vision AI.
Bosch’s concept of Manufacturing Co-Intelligence® also reflects a broader trend in industrial AI. Rather than replacing skilled workers, AI systems are being designed to augment engineers and operators by providing faster analysis and operational recommendations.
As AI systems continue to evolve, engineers will increasingly collaborate with digital agents that monitor production environments continuously. This shift allows industrial experts to focus on strategic optimisation and process improvement while AI systems manage complex data analysis in real time.
As adoption accelerates, Agentic AI in manufacturing is likely to become a defining capability for future smart factory operations.
FAQ: Agentic AI in Manufacturing
What is Agentic AI in manufacturing?
Agentic AI refers to artificial intelligence systems composed of multiple intelligent software agents that analyse factory data and recommend actions to optimise manufacturing operations.
How is Agentic AI different from traditional industrial AI?
Traditional industrial AI focuses on analysing machine data and generating insights. Agentic AI coordinates actions across machines, production systems, and operational processes.
What benefits can Agentic AI deliver to factories?
Agentic AI can reduce downtime, improve production planning, accelerate problem diagnosis, and support engineers with faster data-driven decision making.
Why are semantic data structures important for industrial AI?
Semantic data structures create a shared data language across machines and software systems. This enables AI systems to understand relationships between production datasets.
Will AI replace workers in factories?
Most industrial AI strategies are designed to augment human expertise rather than replace it. Engineers remain responsible for oversight and decision making.
How could AI change the role of engineers?
Engineers will increasingly supervise AI systems, analyse insights generated by digital agents, and focus on improving long-term manufacturing performance.
