What I witnessed firsthand about manufacturing DX was already beyond the realm of "consideration" and had reached the stage of "implementation."
The most prominent theme at Hannover Messe 2026 wasn't "how far AI can go."This year's theme, "Think Tech Forward," wasn't about pushing technology forward, but rather about a commitment to continuously moving forward in the face of challenges. Even in an era where agentic AI is the norm, the final decision and responsibility will rest with humans—this "Human in the Loop" approach was the practical solution repeatedly demonstrated at the event.
Looking back at Hannover Messe over the past few years, the focus of discussions on manufacturing DX has shifted slightly each year. 2024 was on sustainability and standards, 2025 on implementation spanning IT and OT, and 2026 marked the beginning of the stage where AI is continuously implemented as part of real-world operations. In this article, based on messages received at the event, we will organize why Mendix was at the center of the discussions in 2026 within this context.
Hannover Messe 2026: The Current State of DX in Manufacturing
From concept to implementation and operation—the shift in the focus of DX over three years
First, let me briefly summarize the changes over the past three years to give you an overall picture.
In2024, "sustainable manufacturing" and carbon footprint reduction were the main topics. The de jure standard strategy, led by Europe (especially Germany), was strongly promoted, as exemplified by the Battery Passport, and generative AI also emerged, but the focus was more on technology introductions and presentations of future visions than on use cases. Overall, it can be said that it was a year that set direction and ideologies.
By2025, the theme shifted to "industrial transformation," with the practical application of AI and the evolution of edge devices coming to the forefront. As AI /data entered the implementation phase, and application implementation and data integration progressed on the edge side, the collaboration between IT and OT accelerated dramatically. While the specifics of data/ AI utilization from the OT side increased, the resolution of the challenges and realities of OT sites from the IT side also greatly improved.
At the same time, this year also clearly highlighted the difficulties in handling sensing data. Environmental factors such as machine differences and noise, resonance due to mounting location, differences in mounting methods, and coordination with metadata—challenges in "data acquisition itself" that were difficult to see from an IT perspective—were strongly recognized, and it is symbolic that attention to virtual sensing increased. Compared to 2024, this year clearly saw a shift in focus to implementation in the real world.
Building on that trend, it's no exaggeration to say that 2026 was entirely dominated by Agentic AI.
What has emerged is a "complete AI-centric" landscape, but the reality is far from simple.
At the 2026 conference, agentic AI, digital twins, and low-code development were being discussed everywhere. However, what I strongly felt while walking around the venue was that "AI has advanced, so it's not necessarily easier to understand." There were a great many exhibits on generative AI and agents, and to be honest, there were many instances where the superficial messages seemed similar.
"Where will we differentiate ourselves?" "Is it really usable in real-world situations?"
These questions quietly permeated the entire venue without clear answers. While DX is certainly moving forward, there was hardly any sense that it had become"easier."
Siemens' "end-to-end" approach and what's behind it
At Hall 27, Siemens presented a vision of a digital world that seamlessly connects design, manufacturing, and operations. It assumed the integrated handling of IT, OT, data, and AI, and DX was no longer seen as a special initiative, but rather as an essential infrastructure for manufacturing.
On the other hand, from every detail of the booth presentation, it was clear that there are still extremely difficult challenges remaining, such as dealing with vast and fragmented data, organizing correlations, and making decisions across domains.
A calm perspective on physical AI and robots
While physical AI /humanoids, which garnered attention at CES, were also exhibited by various companies at Hannover, their novelty was limited. Many felt that the practical application of humanoids was "still a long way off," and compared to digital generative AI, it left a strong impression that it was difficult to envision their practical applications.
Companies that have already begun pilot programs, in particular, have expressed the opinion that while directly comparing robots to human workers can easily lead to disappointment, the real challenge lies in the fact that they are not yet capable of developing workflows and roadmaps that are based on collaboration with robots.
Why is Mendix at the heart of manufacturing DX?
Considering these developments over the past three years and the general atmosphere of 2026, it becomes clear why Mendix had such a strong presence. The biggest reason is that it was designed with the premise that "humans will ultimately be involved (Human in the Loop)," even in the age of agentic AI.
Mendix didn't present a world view where everything was left to AI; instead, it clearly demonstrated a structure where AI made judgments and suggestions, which were then verified, approved, and ultimately taken responsibility by humans. Agents were assigned a Confidence Level, and if the threshold was not met, the process was not executed automatically and was returned to a human. The human's judgment was then reflected in the next learning process. This design philosophy was highly praised for being aligned with the realities of manufacturing environments.
As agentic AI evolves from "recording and suggesting" to "executing," the importance of human governance becomes even greater. When AI autonomously performs tasks—especially in situations involving physical impact such as operating equipment or placing orders—the final judgment and responsibility for the results must rest with humans. Mendix 's Human in the Loop design philosophy directly addresses this governance requirement.
While each company used different terminology, such as 3-layer or 4-layer architectures, the common message across all companies was that operating Agentic AI requires an architecture where the application layer, data hub layer (AI fabric layer), and data source layer are organically connected.
From comprehensive manufacturing domain vendors like Siemens, to platform providers like Microsoft and AWS, and SaaS vendors like SAP and Salesforce, the premise that "data sources cannot be handled by a single company" was shared, and the importance of the data hub layer was emphasized.
In that sense, this year's Hannover Messe could be described as a "data hub battle" revolving around how to best utilize agentic AI.
Furthermore, the strong attention given to Knowledge Graph / Graph DB as a technology supporting that data hub layer naturally connected with the discussions surrounding Mendix.
What we learned from Macnica 's booth
Macnica 's participation in the Mendix booth as a Japanese partner stemmed from a clear intention to "bring the realities of Japanese manufacturing sites into a global context." The exhibit went beyond mere app development, presenting practical solutions that delved into "how to make business operations work and how to execute them." At the heart of this was a demonstration of "Dxter," an AI agent developed by Macnica and Macnica partner, Orangeleaf.
Dxter seamlessly connects multiple core systems such as Teamcenter and SAP, integrating a series of business processes from problem recognition and analysis to parts selection, vendor identification, RFQ creation, email sending, and work order generation. Traditionally, these tasks were fragmented manual processes that took hours to weeks due to multiple handovers and decision-making processes. In contrast, Dxter automatically collects and organizes necessary information, generates the documents required for decision-making, and carries out the entire process from start to finish. As a result, the entire process is compressed into minutes, demonstrating "automation of execution itself."
Even more importantly, this system is not merely an automation demonstration, but also includes "designing the roles of humans and AI." Dxter has not only a fully autonomous mode but also a hybrid mode where humans review the process, and it is based on a realistic operational design that considers "how much to leave to AI and what to handle with humans." This strongly indicates that it is not a vision for the future, but a "business design that can be implemented immediately."
Following this demonstration, many attendees raised highly implementation- and operational questions, such as "How is the data being connected?" and "Who is involved on the ground and how?" This was common not only among Japanese participants but also among international attendees, confirming that data integration and human involvement design are essential global concerns.
Furthermore, the context in which the Japanese use case was presented reinforced this context. Japanese manufacturing sites face a combination of challenges: ① loss of tacit knowledge due to the retirement of veteran employees, ② labor shortages, and ③ increased workload due to business growth. In response to this, an approach was presented to build a "hybrid workforce" by integrating AI agents as "digital workers" into organizations and having them collaborate with humans.
As the foundation for this, Mendix provides ① Secure (a secure execution environment), ② Open (freedom to select generated AI), and ③ Connect (system integration including ERP and legacy systems), playing a role not merely in application development, but in integrating and enabling the entire business process.
Through these exhibits, it was evident that Macnica and Orangeleaf 's efforts went beyond simply showcasing tools and technologies, demonstrating how they address on-site challenges and support the entire business process. The essence of this booth lay in its ability to connect a realistic understanding of manufacturing challenges with cutting-edge global technologies (Mendix and Aggent AI), presenting "concept" and "execution" as a unified whole.
To those in charge of DX in Japanese manufacturing: Questions brought back from the field.
What I realized again through Hannover Messe 2026 is the difference in the approach to promoting DX between Japan and Europe. In Europe, the IT department is positioned not as the "builders," but as the "connectors and supporters," and there is a widely shared premise that the frontline staff will take the lead in continuously driving improvements.
As a result, instead of finalizing all requirements before taking action, the approach of "deciding as you go" in DX—starting to run and making adjustments as you use it—is functioning as a realistic option. This may be attributed not so much to a difference in technical level, but rather to a difference in the division of roles and the structure of decision-making.
For those in charge of DX in Japanese manufacturing, what's important is not so much how to introduce the latest technologies,
- To what extent should humans make decisions, and to what extent should they be entrusted to AI?
- At what stage will the on-site staff remain involved?
The key is whether you can ask questions that are based on such operational assumptions from an early stage. Agentic AI and Human in the Loop are not the goal itself, but merely means to be chosen as you confront those questions.
What was shared at Hannover Messe 2026 was the very reality that each company is still searching for the right way to proceed. I hope that the trends and questions that emerged here will serve as a reference when you consider your own situation.
in conclusion
The digital transformation (DX) in the manufacturing industry is shifting from an era of "not knowing where to start" to an era of "how to keep it running." As a partner that supports implementation rooted in the manufacturing sites of Japanese companies through Mendix, Macnica will work with you to advance your DX into the "implementation phase."
For information on how to use Mendix, demo experiences, and implementation consultations, please feel free to contact Macnica.