"Automating automation" has become a product. How will the Siemens Eigen Engineering Agent change the work of control engineers?
Hannover Messe 2026 is one of the world's largest industrial technology trade fairs held in Germany. At the 2026 venue, the unveiling of humanoids—robots with a form close to humans—was particularly prominent.
For example, Siemens, a major German industrial equipment and automation company, announced a case study at its Erlangen electronics factory where the HMND 01 Alpha from Humanoid performed logistics tasks. Humanoid is a company that develops industrial humanoid robots, and the HMND 01 Alpha is their wheeled robot. It was news that was easy to understand visually, conveying that robots are beginning to move within factories.
However, another announcement made by Siemens on April 20th might have a greater impact on the practical work of factory DX. That is the Eigen Engineering Agent.
The Eigen Engineering Agent is an industrial AI agent used within TIA Portal, Siemens' automation engineering environment. TIA Portal is Siemens' integrated engineering environment for designing and configuring PLCs, HMIs, and more. PLCs are devices that control factory equipment, and HMIs are screens that allow people to view and operate the status of equipment.
Siemens has embedded an AI agent into this TIA Portal. The scope includes PLC code generation, HMI screen creation, device configuration, and searching for or explaining existing projects. The target users are the more than 600,000 TIA Portal users worldwide.
Rainer Brehm of Siemens Digital Industries describes this as "automating automation." In Japanese, this means "automating the automation."
Until now, the main theme has been how to automate the work of people and robots. This time, the work of the engineers who design that automation is beginning to become the target of AI.
From the perspective of factory DX practice, this might be a more significant announcement. The reason is that it is not about the performance of a single robot, but about directly touching the engineering work of 600,000 people.
From "AI that proposes" to "AI that executes"
Industrial AI until now has been centered more on the "copilot" type. When you ask a question in natural language, it provides code suggestions, searches for content in specifications, or shows candidates for settings. From that point on, human engineers had to make judgments, make corrections, and incorporate them into tools.
The Eigen Engineering Agent aims one step further. According to Siemens' explanation, it is positioned as one of the industrial AI agents that has entered the commercial provision phase.
Vasi Philomin, EVP of Siemens Data and AI, explains this change as moving to a stage where AI does not just offer advice from the sidelines, but breaks down tasks, executes them, and organizes them into a form that can be verified within actual design tools.
The "autonomous execution" referred to here does not mean leaving final judgments to AI without human review. It means that AI takes on the responsibility of breaking down tasks, executing them, verifying them, and producing them as reviewable deliverables. Final confirmation and adoption decisions remain with human engineers. This difference is not small.
Copilot-type AI makes an engineer's work faster. On the other hand, agent-type AI takes on parts of the work collectively. The former is a tool to "speed up hands," and the latter is a tool to "reduce the effort itself."
The work of a control engineer includes not only creative judgment but also a lot of tedious and time-consuming tasks. Searching for necessary blocks from past projects. Fixing similar code. Standardizing naming and styles. Adjusting screen components. Changing parameters in bulk. Checking documents. It is precisely this area that the Eigen Engineering Agent has begun to enter.
The targets are "writing, fixing, searching, standardizing, and interpreting"
Looking at the Siemens product page, the tasks that the Eigen Engineering Agent targets are quite specific. It might be easier to understand them as on-site chores rather than listing difficult function names.
For example, there is the task of writing control programs that describe how equipment should move. Writing similar processes while looking at past code, creating logic for testing, and fixing it if errors occur. AI supports these tasks.
The task of creating equipment operation screens is also a target. This includes support for creating screens that operators look at, screens that display equipment status, and processes that control the movement of buttons and displays. It also includes the task of migrating screens created with old mechanisms to new ones.
Furthermore, there is the task of standardizing settings. Changing settings across multiple devices or parts in bulk, checking if they comply with company rules, and creating necessary documents. These tasks are not conspicuous, but they are parts that take time and are prone to errors when done by hand.
Another major aspect is the task of deciphering past projects. When you ask, "Show me the part that controls this process," it becomes possible to find the relevant control blocks and structures. For those who join a project midway, this helps in understanding the existing design.
Finally, there is support for translation, explanation, and troubleshooting. It can translate on-screen text into multiple languages, explain technical information, and provide confirmation procedures when problems occur. This is intended to reduce the time spent on research and verification tasks.
What is happening here is that among the tasks of a control engineer—"writing," "fixing," "organizing," "deciphering," and "searching"—are gradually becoming the target of AI agents.
AI is beginning to take on, little by little, the tedious tasks that pile up before a control engineer can reach a decision. And until now, that effort has not been very visible.
The significance of reducing new employee training from "weeks" to "days"
Among the cases introduced by Siemens, the most easy-to-understand one is the training of new employees at an automotive line builder. An automotive line builder is a company that designs and constructs production lines for automobile factories.
Conventionally, it took several weeks for a new engineer to join a project and become a productive member of the team. The reason is not simply that they need to learn how to use TIA Portal. It is because they need to decipher the structure of existing projects, the relationships between control blocks, the connections between screens and devices, and past design intentions.
Control projects often contain a lot of context that cannot be understood from drawings and specifications alone. Why is this block named this way? Why does only this signal have exception handling? Which modification led to the current form? Just searching for such information takes a considerable amount of time.
Using the Eigen Engineering Agent, a new engineer can ask the project questions in natural language. If they ask, "Show me all the blocks that control Station 3," they can immediately check the relevant information. Siemens explains that as a result, the onboarding period has been shortened from weeks to days.
This is not just a shortening of the training period. In control engineering, the fact that only a limited number of people can decipher existing projects is often a bottleneck. You can't understand it unless you ask a veteran. You're afraid to fix it unless you're someone who has worked on past projects. Even when a new person joins, they spend their first few weeks on investigation and understanding.
If AI can support that deciphering process, it becomes easier for new people to enter projects that have become highly specialized. This is also a matter related to labor shortages and the issue of technical succession.
On the other hand, there are points to be aware of. When AI becomes able to explain existing projects, onboarding becomes faster. However, it does not automatically inherit all the tacit knowledge, such as why that design was chosen, what troubles occurred in the past, and where on-site judgments were incorporated.
The shorter the time spent deciphering, the more we need to be conscious of what humans should understand and design accordingly.
General availability after being tested by over 100 companies in 19 countries
The Eigen Engineering Agent is not just a concept for announcement. Siemens explains that it has begun general availability after pilot implementations in over 100 companies across 19 countries. This means that before the official sale, it was used by actual companies to confirm its effectiveness and challenges.
Some company names have been disclosed. China's CASMT is a company that handles high-performance production lines for new energy vehicles. The company uses the Eigen Engineering Agent for device configuration, code generation, and HMI visualization on its EMB (electromechanical brake) lines. By changing to a conversational workflow, handovers between experts have decreased, leading to debugging and shorter delivery times.
Austria's ANDRITZ Metals is a company that handles metal processing equipment and the like. The company utilizes it for code generation, document generation, and problem-solving within TIA Portal.
Prism Systems in the United States uses it for the generation and import of control programs, and has confirmed a reduction in the time required for those tasks.
Looking at this lineup, the target is not limited to a specific industry. In multiple fields such as automotive, metal processing, and system integration, it is being used in a way that commonly reduces the "effort of control engineering."
What Siemens is aiming for is not individual convenient features, but the integration of AI into the daily work environment itself, which is TIA Portal. This is slightly different from the idea of launching and using AI tools separately.
There is an AI agent within the environment that engineers open every day. This difference affects both the ease of adoption and the ease of long-term integration.
Factory DX is progressing in areas separate from the AI-ification of robots
In recent discussions about factory DX and Physical AI, attention tends to focus on the robots themselves. What can a humanoid hold? How dexterously can it move? Can it work in the same space as humans? Such topics are easy to understand and make for good video footage.
However, factory automation is not built on robots alone. To operate robots and equipment, there are PLCs, HMIs, sensors, drives, networks, and the control design that connects them all. What really takes time on the factory floor is not the flashy demos, but the design, configuration, modification, and handover processes.
The Eigen Engineering Agent is an AI that has entered that backend space. It targets the engineering work required to operate robots and equipment. Its significance lies in shortening the time spent on control programs, operation screens, settings, and interpreting past designs, making it easier to continue using in the field.
When people talk about using AI in factories, applications like anomaly detection, image inspection, predictive maintenance, and robot control come to mind first. Of course, those are important too. However, before and after those tasks, there is the work of the people who design, configure, connect, and repair. If that work gets bottlenecked, no matter how good the AI or robots are, they will not take root on the factory floor. Siemens is beginning to target the engineering work itself, which is where time is often lost on the shop floor, with AI.
Schneider-style "selection" and Siemens-style "work automation"
Previously, regarding the AI utilization of Schneider Electric, a major French electrical and industrial automation company, I summarized that the ability to determine "which AI to put into production" is crucial to avoid getting stuck in PoC. PoC refers to the stage of trial implementation to confirm effectiveness. Many companies face the challenge of being unable to move from this small-scale verification to full-scale operation, often getting stuck there.
The Schneider Electric story was about the organizational side: which areas should AI be introduced into, and how should it be selected? On the other hand, this Siemens story is about the tool side: how to reduce the engineering work after the selection has been made.
These two are not contradictory. First, determine which operations or equipment should have AI introduced. Next, use AI agents to shorten the actual work of introduction, configuration, and modification. When both are in place, the distance from small-scale verification to full-scale operation is significantly shortened.
In factory DX, even if the ideas for implementation are clear, there is often a shortage of manpower to actually design, configure, and modify. Control engineering, in particular, is heavily context-dependent for each piece of equipment, and both handovers and modifications take time. The Eigen Engineering Agent is a product that reduces the burden of engineering work that tends to balloon during such implementation phases.
Rockwell is looking in the same direction
This movement is not limited to Siemens.
Rockwell Automation also demonstrated an AI-orchestrated Factory System Design at Hannover Messe 2026. Rockwell Automation is a major US industrial automation company. In the company's announcement, a direction was shown to support factory system design with AI by combining Emulate3D, which reproduces factory movements in a virtual space, Visual Studio Code, GitHub Copilot, and FactoryTalk Design Studio, which supports control design.
However, the maturity level here is different from Siemens. The Siemens Eigen Engineering Agent has already begun general availability. On the other hand, the Rockwell announcement is positioned as a demo exhibit. Therefore, it is not accurate to view both companies as being at the same "shipped" stage. Even so, the direction is similar.
Companies with control platforms are beginning to incorporate AI into the very work of PLC engineering and factory design. This indicates that the axis of competition in the automation industry is beginning to shift slightly.
Until now, the major axes of competition were the performance of control equipment, the ease of use of software, and stability on the factory floor. From now on, how much engineering work can be shortened, how easy it is to understand past control programs and screen settings, and how quickly new or mid-career members can join in will also become important differentiators.
The work that remains after "automating automation"
The name Eigen, used by Siemens, means "own" or "inherent" in German. For those in science and engineering, it is also a term that brings to mind "eigenvalue." Siemens explains the name as representing a stable intelligence rooted in industrial domain knowledge, even amidst the rapidly changing landscape of AI.
As a name, it is quite well-crafted. However, what is truly important is not the name. Out of the more than 600,000 TIA Portal users, how many will begin using the Eigen Engineering Agent on a daily basis? Which tasks will be entrusted to it, and which judgments will be left to humans? How will the code and settings created by AI be reviewed and quality-assured? This is the next point of discussion.
In factory control, mistakes can lead to equipment shutdowns or quality defects. Even if AI can speed up tasks, it cannot simply be accepted on the shop floor as is. Review, verification, standardization, and a clear division of responsibility are required.
In other words, the role of a control engineer will shift: the proportion of manual work will decrease, while the importance of evaluating what the AI has created and refining it into a form that can be used on the shop floor will increase.
The time spent writing code from scratch may decrease. The time spent searching through existing projects may also decrease. The work of standardizing screens and settings may also become shorter.
Instead, the ability to evaluate what the AI produces, the ability to make judgments based on equipment and process constraints, and the ability to translate these into a form that can be safely used on the shop floor will become more important.
"Automating automation" is no longer just a slogan; it has begun to become a product. The next question is what control engineers will let go of and what they will take on anew in their work.
AI colleagues are already starting to take their seats inside the TIA Portal. The transformation of factory DX is also beginning to advance from the desks of those who operate robots and equipment.
Articles related to this theme
・Siemens brings AI to design, Bosch brings it to the shop floor | Two entry points for factory AI seen at Hannover Messe 2026
・AI is not just answering, it has started giving instructions to control systems | A new contact point for factory AI shown by Beckhoff
・Which cloud will factory AI run on? The differences between Schneider×Microsoft and Rockwell×AWS seen at Hannover Messe 2026
How to apply this theme to your own company
For those who want to organize which tasks to consider for factory AI, I have published a paid article comparing five companies: Siemens, Bosch, Beckhoff, Schneider Electric, and EDAG. I have organized perspectives for internal consideration based on each implementation starting point: design/engineering processes, active shop floors, control layers, and cloud infrastructure.
Where should you start thinking about AI for manufacturing? A map of implementation starting points for your company, seen through Siemens, Bosch, Beckhoff, Schneider, and EDAG
Regarding consultations
For executives and corporate planners at small and medium-sized manufacturing companies considering the introduction of AI for manufacturing or factory AI, I provide assistance as a personal activity, ranging from "organizing issues" before consulting with SIers or AI vendors to creating prototypes for small-scale trials.
Which tasks should AI be applied to? How much should be outsourced, and what should remain in-house? How should the conditions for moving from a small-scale trial to full-scale investment be set? Can you create a structure where you can continue to make your own judgments without becoming overly dependent on external support after implementation? I will help you organize these points.
I do not stop at just organizing the issues. If necessary, I will prepare a working prototype or trial environment within a few weeks so that you can decide on the next steps while looking at the actual product.
Please note that these consultations are based on publicly available information and general business organization concepts. I do not provide advice related to your organization's specific business, clients, or technical information.
Service details, pricing, and the inquiry form are summarized here.
https://chiteki-tankyu-hub.com/
The initial 30-minute online consultation is free. Feel free to reach out even at the stage where you are just wondering where to start. For paid brainstorming sessions, I will summarize the points organized during the session into a single-page 'memo of issues to check before implementation' and provide it to you.
You can also reach me via the inquiry form on this note or by DM on X.
Thank you for reading until the end.
In this note, I organize trends in factory AI, robots, and smart factories by connecting management, technology, and on-site implementation.
I don't just introduce individual news; I also track which companies enter factory DX from which entry points and where the difficulties in on-site implementation arise.
If you would like to continue reading about this theme in the future, I would be happy if you could follow me.
Sources and Reference Links
Siemens AG｜"Siemens launches the Eigen Engineering Agent, bringing purpose-built AI to industrial automation engineering"｜https://press.siemens.com/global/en/pressrelease/siemens-launches-eigen-engineering-agent-bringing-purpose-built-ai-industrial
Siemens AG｜"Siemens brings AI to the physical world with Eigen Engineering Agent"｜https://press.siemens.com/global/en/pressrelease/siemens-brings-ai-physical-world-eigen-engineering-agent
Siemens｜"Eigen Engineering Agent - TIA Portal"｜https://www.siemens.com/en-us/products/tia-portal/eigen-engineering-agent/
Manufacturing Digital｜"How Siemens Deployed One of the First Industrial AI Agents"｜https://manufacturingdigital.com/articles/how-siemens-ai-engineering-agent-eigen-works
Robotics & Automation News｜"Siemens launches AI engineering agent to automate PLC coding and industrial workflows"｜https://roboticsandautomationnews.com/2026/04/20/siemens-launches-ai-engineering-agent-to-automate-plc-coding-and-industrial-workflows/100758/
AI News｜"Siemens introduces AI system for automation engineering"｜https://www.artificialintelligence-news.com/news/siemens-ai-automation-engineering-workflows/
Siemens AG｜"Siemens and Humanoid bring Physical AI to the factory floor"｜https://press.siemens.com/global/en/pressrelease/siemens-and-humanoid-bring-physical-ai-factory-floor-deploying-humanoids-industrial
Rockwell Automation | "Rockwell Automation to Demonstrate AI-Orchestrated Factory System Design at Hannover Messe 2026" | https://www.rockwellautomation.com/en-nl/company/news/press-releases/ai-orchestrated-factory-design-at-hannover-messe.html