Siemens has introduced an AI agent that can plan, execute and validate industrial automation engineering. The Eigen Engineering Agent works directly inside the company’s TIA Portal, where it can access the structures, logic and component relationships of a real project.
The system targets programmable logic controller code, human-machine interfaces and device configuration. It breaks requirements into steps, generates work, evaluates the result against predefined criteria and iterates before presenting it to an engineer.
Siemens says tasks can run two to five times faster than manual workflows. More than 100 companies across 19 countries have participated in pilots, including ANDRITZ Metals, CASMT and Prism Systems.
Project context makes automation useful
Generic coding assistants struggle with industrial environments because every plant contains specific equipment, control logic and engineering conventions. Legacy systems may also be poorly documented.
Integration with TIA Portal lets Eigen reference the project’s hierarchy, dependencies and existing configurations. It can generate structured control language aligned with the actual system rather than producing isolated code that an engineer must translate manually.
Prism Systems used the agent to create and import SCL code, reducing execution time. CASMT applied it to device configuration, code generation and HMI visualisation for production-line development, reporting fewer specialist handoffs and shorter delivery timelines.
The agent’s self-correction loop is important. Industrial correctness cannot be judged by whether code compiles alone; outputs must satisfy project requirements and preserve relationships across devices.
Workforce shortages increase the value of assistance
Industry estimates cited by the source suggest a global shortage of up to seven million manufacturing workers by 2030, with roughly one in five engineering roles unfilled in some sectors.
AI cannot manufacture experienced automation engineers instantly, but it can remove repetitive configuration and give specialists more time for architecture, safety and exception handling. It may also help less experienced staff navigate existing project structures.
TIA Portal has more than 600,000 users, giving Siemens a large installed base for distribution. Eigen is available through the Xcelerator portfolio, avoiding the integration burden of a separate general-purpose agent.
Skills remain a constraint. Manufacturers need people who understand both industrial processes and the limits of AI-generated work. An engineer must be able to recognise when the system has met a formal target but violated an operational assumption not captured in the project data.
Industrial autonomy requires verified boundaries
PLC code can control physical equipment, so review and change management are non-negotiable. Generated logic should run in simulation or a safe test environment before reaching production, with clear records of what the agent changed and why.
Data quality is another barrier. Manufacturers may possess large volumes of operational information without consistent context. An agent drawing from incomplete device definitions or undocumented exceptions can accelerate the wrong design.
Siemens has invested €1 billion in industrial AI and reports more than 1,500 AI specialists and over 2,000 related patent families. Eigen is positioned to extend beyond automation engineering into other parts of the industrial value chain.
That expansion should follow evidence from bounded workflows. Code generation and device configuration offer measurable baselines in engineering time, defects and rework. Broader authority will require stronger validation and safety cases.
The agent illustrates where industrial AI may deliver near-term value: inside the tools engineers already use, grounded in project-specific information and required to pass defined checks.
It does not remove engineering judgement. It compresses the path from requirement to reviewed implementation. If Siemens can maintain traceability and reliability as the agent’s scope grows, automation engineers may spend less time translating routine specifications and more time ensuring that increasingly intelligent factories behave safely in the real world. That is a meaningful operational gain.
