From agile to autonomous – building adaptive production with digital twins and industrial AI
Manufacturing has always been defined by constraints: material limits, workforce capacity, supply chain reliability. For decades, the goal was to optimize within those constraints. The most competitive manufacturers are now doing something different, they are redesigning the system that generates constraints in the first place.
This is the core discussion Zvi Feuer, Senior Vice President of Digital Manufacturing, and Mark Hindsbo, Head of Operations Software at Siemens Digital Industries Software, made in their recent keynote on building adaptive production with the digital twin and industrial AI. The path from where most manufacturers are to where they need to be is neither simple nor instantaneous. But it is, for the first time, within reach.
Why “agile” is no longer enough
Agility is reactive by design. It assumes the disruption has already happened, the demand signal has already shifted, the supply chain has already broken. An agile factory responds well. An adaptive factory anticipates. An autonomous factory acts, often before a human operator registers that a decision needs to be made. The distinction matters. Think about it from this lens:
Agile: fast reconfiguration in response to change. The cognitive load still sits with human planners and operators.
Adaptive: real-time data and predictive analytics drive proactive adjustment. The system does more reasoning; humans provide oversight.
Autonomous: the system senses, decides, and acts with minimal human intervention. The human role shifts from operator to architect.
Most manufacturers sit somewhere between agile and adaptive. The barrier to moving further is rarely technology. It is a data continuity gap and closing it requires rethinking how digital information flows across the entire enterprise.
Unifying manufacturing engineering and production operations
Manufacturers are grappling with extreme complexity, from macroeconomic volatility and geopolitical uncertainty to rapid advances in digital technologies like artificial intelligence. These forces are demanding more than just incremental improvements; they require a fundamental shift in how production systems are designed, operated, optimized, and sustained. The traditional approach of isolated applications and device-driven automation is no longer sufficient. Success now hinges on embracing data-centric factories and unifying intelligence across the entire production lifecycle.
This is precisely why Siemens, with its extensive portfolio of software and automation solutions, is uniquely positioned to address this industry need. Siemens’ comprehensive suite of solutions is designed to provide a unified digital thread that seamlessly bridges the digital (ET/IT) and real (OT) worlds.
Hear from Mark Hindsbo, Head of Operations Software at Siemens Digital Industries Software, about how Siemens is helping bridge the digital (ET/IT) and real (OT) worlds.
Unlike competitors who often optimize individual functions, Siemens delivers system-level transformation for the entire production process, from production engineering to operations and service. By connecting data across the entire production lifecycle, our solutions enable companies to move quickly, automate and capture best practices to augment worker shortages, and drive autonomous operations globally.
This modular and integrated set of software and automation capabilities provides the backbone for the next generation of manufacturing, where data drives production capabilities and decisions. It positions organizations to progress from optimized execution toward autonomous, self-optimizing production environments where human teams are amplified by trusted, industrial-scale AI. Siemens, with its deep industrial domain expertise and comprehensive digital thread, is a long-term strategic collaborator, supporting customers at every stage of their transformation journey.
For example, from the manufacturing engineering perspective, this involves using integrated, computer-based systems with simulation, 3D visualization, analytics, and collaboration tools to simultaneously define both the product and the manufacturing process. This allows industries to design entire manufacturing processes digitally, fostering seamless collaboration between engineers and designers, from digital manufacturing solutions like Tecnomatix, NX for Manufacturing, Opcenter, and Teamcenter Manufacturing, and shop floor applications and equipment. This exchange of product-related information between design and manufacturing groups is crucial for achieving time-to-market and production volume goals, while also reducing costly downstream changes.
To connect engineering with production operations, this includes powerful solutions like gPROMS for process modeling, COMOS for plant engineering, SIMIT for the simulation of automation functions (DCS and SCADA) and XHQ, which provides operations intelligence for several global oil and gas industry supermajors. Siemens has been building smart manufacturing capabilities for decades, from CAM and process simulation on the engineering side, to SCADA, MES, and MOM on the operations side. Few companies can claim that depth across both worlds.
That is why Siemens is responding by bringing its portfolio closer together across software and automation, unifying in a single product organization all the capabilities needed to design and operate the factories and plants of today and tomorrow. The foundation is our AI-native, agentic enterprise strategy. Underpinning it all is continued investment in the digital thread connecting product design to factory automation: an end-to-end capability that only Siemens can deliver at this scale.
What makes this more than a portfolio story is how we are building it, including inside our own factories. Three principles define the approach:
- AI must operate on the current state of the factory. Not a historical snapshot, not a periodic sync, the live state. Only then can AI move from analysis to action, making decisions grounded in what is actually happening on the shop floor right now.
- Action requires Six Sigma confidence. In manufacturing, an AI decision made on uncertain data isn’t neutral, it is a liability. That level of confidence demands a deterministic digital twin running in operation, not just in planning.
- Execution must be auditable. AI in industrial environments isn’t a black box. Every action needs to run through industry-grade, tested processes that can be reviewed and trusted by the engineers and operators who are ultimately responsible for what comes off the production line.
This is what Siemens manufacturing engineering and production operations brings together: live operational data, a deterministic digital twin, and an auditable execution layer. Together, they form the foundation for AI that does more than observe the factory; it can reliably run it.
The digital twin: not a model, a mirror
A digital twin is only as valuable as its connection to reality. Siemens’ approach centers on an “always-live twin,” a continuously updated replica of the entire production system, from product design and manufacturing process to the factory floor.
The architecture makes this continuity structural. The engineering bill of materials (EBOM) automatically generates the manufacturing bill of materials (MBOM), which drives the bill of process (BOP). The twin of the product generates the twin of the manufacturing process, which generates the twin of the production system. When a design changes, effects propagate automatically, engineers simulate the impact before anything physical is touched. Design issues caught in the digital twin cost less than 1% of what they cost to resolve in production. This is the compounding logic of the live twin: it does not just improve one metric, it lowers the cost of change itself.
Industrial AI: the intelligence that activates the twin
A digital twin without AI is a sophisticated monitoring tool. With AI, it becomes a decision-making system, the “production brain” trained on product and production design data, grounded in manufacturing realities.
Hear again from Mark Hindsbo, Head of Operations Software at Siemens Digital Industries Software, about how Siemens is helping manufacturers accelerate their success using AI-powered technologies.
Four scenarios stand out to represent this:
- Brownfield integration. Most manufacturers are not building greenfield. Their operational intelligence lives in legacy MES platforms, process historians, P&ID diagrams, and DCS configurations that rarely talk to each other. Siemens’ agentic AI can help companies unify fragmented data, contextualizing it into a common language, orchestrate decisions across systems and close the loop between insight and action to eventually construct a comprehensive digital twin of existing facilities, meaning the path to autonomy is open to manufacturers with decades of embedded operational history, not just those starting fresh.
- Autonomous scheduling. Siemens has deployed autonomous scheduling in its own factories, not as a pilot, but as standard operations. The system continuously re-optimizes in response to real-time changes in demand, machine availability, and supply. What once required skilled planners working with incomplete data now happens algorithmically, at a speed no human team can match.
- Predictive quality and maintenance. AI-driven quality control moves defect detection upstream, identifying process conditions likely to cause issues before they appear in finished goods. Predictive maintenance shifts from fixed schedules to condition-based intervention, acting before a machine fails rather than after.
- AI agents as automation primitives. The most forward-looking shift: AI agents replacing deterministic coded logic. A “Brewmaster Agent” that autonomously manages the complex variability of fermentation is one example. The principle extends to any process where variability and expert judgment have historically made automation difficult.
The scaling problem most companies don’t solve
The manufacturing industry has no shortage of successful AI pilots. It has a significant shortage of pilots that became enterprise programs. The failure mode is predictable: a motivated team produces real results in a bounded scope, then momentum stalls when leadership tries to replicate success at scale. Pilots operate outside normal constraints – dedicated resources, senior attention, permission to work around legacy systems. Scaling requires integrating with those systems.
Three things differentiate organizations that make it through:
- Data continuity. Pilots run on curated data sets. Scale requires a robust AI data fabric ensuring consistent, high-quality data flows continuously from operational systems into the models that depend on it. Unglamorous work. Non-negotiable.
- Operating model evolution. Autonomous production changes what people do. Engineers shift from writing code to governing AI agents. Planners shift from generating schedules to auditing algorithmically generated ones. Organizations that treat this as a technology project rather than an operating model transformation find that the technology works and the organization does not.
- Clear decision authorities. Autonomous systems need explicit boundaries: what the system decides alone, what requires human confirmation, what always requires human judgment. Over-escalating loses the efficiency gains; under-escalating creates unmanaged risk.
Hear from Zvi Feuer, Senior Vice President of Digital Manufacturing, discussing how data and AI are becoming the “production brain” trained on product and production design data, grounded in manufacturing realities.
For example, this is being done at Siemens’ AI powered Nanjing facility, touted as the most advanced operational sites in the world. The Nanjing factory was recognized for achieving exceptional performance in cost and quality through digital twins and continuous AI-driven transformation.
It was designed, tested, and optimized entirely in the virtual world before a single brick was laid. This approach not only enabled Siemens to construct the factory faster and with outstanding cost-efficiency but also to build it under the toughest pandemic conditions. By combining our global manufacturing expertise with local insight and a digital-first mindset, we continuously optimize every part of the operation, making it one of the most efficient and flexible factories in the world.
What the numbers show:
- Up to 20% increase in throughput
- Up to 30% greater volume flexibility
- Up to 40% improvement in space efficiency
- Up to 50% faster material replenishment
And design issues caught in the digital twin cost less than 1% of what they cost to resolve in production – a figure that captures the compounding value of the live twin better than any throughput metric.
Hear from Zvi Feuer, Senior Vice President of Digital Manufacturing, about Siemens’ own digital native factory, Siemens Numerical Control (Nanjing) Co., Ltd. (SNC) factory, a digital native operation and a living proof point – running its own technology on its own operations:
Building toward autonomous production
The journey toward autonomous production is a sequence of investments, each building the foundation for the next. The right entry point is not the most ambitious application, it is the one that closes the most critical data continuity gap. Manufacturers are asking for something more: capabilities that are more connected, more modular, and increasingly autonomous. Product complexity is rising. Macroeconomic volatility is not going away. And the pace of AI innovation is compressing the window between what is possible and what is expected. That convergence is creating one of the most attractive growth markets in industrial software.
The diagnostic questions are simple: Where do planners spend the most time manually reconciling information that should flow automatically? Where do production disruptions reveal the absence of real-time visibility? Where does a design change in engineering create expensive surprises on the floor?
Those are the places to start. Success there builds credibility and infrastructure to go further – toward AI agents managing complex processes, collaborative robotics with genuine autonomy, and production systems that do not just respond to disruption but prevent it.
The technology is available. The architecture is proven. What separates the manufacturers who lead this transition from those who follow is the consistency with which they pursue the unglamorous work: connecting data, evolving operating models, and expanding the boundaries of what they trust their systems to decide.
