Agentic Industrial AI Will Be All over Hannover Messe; but Will It Meet the Hype?

As over 125,000 professionals are preparing to descend on Hannover Messe for the latest in manufacturing technology innovation, one topic is on the tips of everyone’s tongue, Agentic Artificial Intelligence (AI). Major suppliers, such as ABB, Schneider Electric, Siemens, and Rockwell Automation, have partnered with NVIDIA, Microsoft, and Amazon Web Services (AWS) to bring AI to their offerings. Phoenix Contact and Beckhoff use open platform Open Neural Network Exchange (ONNX) to power their AI platforms.

Moving away from generative copilots, agents are positioned to take over tasks such as Programmable Logic Controller (PLC) code generation, production planning, and plant design. Customers who are historically conservative in their approach to automation are unlikely to be comfortable with their production processes going into the black box of AI. Innovation cycles may be hastening, but adoption is still cautious. Therefore, suppliers must use Hannover Messe to drive home cause for adoption.

Industrial Automation (IA) vendors are moving faster than many customers are ready for. The first virtual PLC was announced 3 years ago by Siemens at Hannover Messe 2023, and deployments of virtual controllers are just now gaining momentum, with the installed base growing from 62,000 in 2025 to 4.8 million in 2035 (a 54.3% Compound Annual Growth Rate (CAGR)) , which still only represents 16% of the overall PLC market (see ABI Research’s Industrial Automation: Software Defined Automation (SDA) market data (MD-SDA-101) for more). Now, IA suppliers are bringing AI agents while customers are still trying to see a Return on Investment (ROI) from 2-year-old generative solutions.

In data-sensitive industries such as aerospace & defense, food & beverage, and life sciences, inputs going into an AI black box are a point of concern because regulations make traceability and transparency a necessity.

Concerns of transparency are layered on top of the mandate that IA vendors such as Siemens, Schneider Electric, and ABB, which are all bringing Agentic AI to their offerings, must prove the value of pilot projects, ensuring that their offerings boost productivity to the degree they promised.

The goal is to deliver models that have fewer parameters and less expensive inference costs, creating Intelligence-as-a-Service (IaaS), whether that be tiered or value-based pricing.

The black box problem cannot be solved in the near term, meaning that vendors such as Siemens, Schneider Electric, and Rockwell Automation must use Hannover Messe as an opportunity to erode the conservative culture among customers. Doing so demands proof of value add with quick time to realization, and highlighting enhanced security surrounding Agentic AI models, creating a framework for trustworthy, transparent, and traceable AI implementations. To effectively drive the adoption of agents, there will be demonstrations of agents using NVIDIA’s NemoClaw platform, which was announced at NVIDIA GTC to drive both private and cost-efficient agents.

Phoenix Contact has taken the latter approach with its MLnext modules on the PLCnext platform, with MLnext Creation allowing for adjustable weightings to the models. While the neural networks of MLnext are a more mature technology than Agentic AI, the approach provides a useful roadmap as Agentic AI moves through its life span.

Hannover Messe will also feature companies positioning with focused Small Language Models (SLMs) to address the Small and Medium Enterprise (SME) market, as over 2,100 industrial AI producers will be attending, with 631 of them targeting automation systems and components. These firms will look to scale down costs associated with AI compute, providing the opportunity to have their less complex models target cost-sensitive customers. These suppliers will leverage the fact that inference costs have historically decreased; however, this may not continue in the future. The trend is dependent on if infrastructure can grow to meet increasing demand and keep inference costs low. IaaS will be enabled to scale down to small and medium manufacturers and scale up to large manufacturers.