A massive transfer of knowledge is imminent in the German manufacturing industry. The baby boomers are retiring, and with them decades of experience. The Hannover Messe 2026 will show for the first time how AI agents take over entire process chains in the factory, far beyond the chatbot.
A shortage of skilled workers costs the industry 49 billion euros per year
The situation is structural: The German Economic Institute puts the annual loss of value added due to a lack of skilled workers at 49 billion euros – across all sectors. At the same time, the experiential knowledge of entire generations is contained in PDF manuals, Excel lists and the minds of individual employees. According to a McKinsey survey, knowledge workers spend an average of 1.8 hours per day searching for information. The problem is exacerbated in manufacturing because technical documentation is often unstructured or out of date.
Specifically, this means: A service technician is faced with an error message on the customer’s machine, looks for the error code in the operating instructions and finds nothing. So he calls his experienced colleague – but he is already retired. Or on another construction site. Or both.
But the problem goes further than the factory floor. It affects purchasing, controlling, customer service and quality assurance. Wherever knowledge is contained in documents and people process this knowledge manually. And along the entire value chain.
Hannover Messe 2026: AI is moving to the center of manufacturing
Deutsche Messe AG is positioning artificial intelligence as an independent exhibition topic for the first time at the Hannover Messe 2026. The mission statement: out of theory and into industrial application. From April 20th to 24th, around 3,500 exhibitors will show how AI actually works in production – not as a research project, but as a productive part of the value chain.
The trend is clearly moving in one direction: away from the isolated chatbot and towards AI process automation with agents who independently analyze documents, compare data and trigger process steps. According to Gartner, around 40 percent of all enterprise applications will contain some form of AI agent by the end of 2026. Or in short: systems that not only respond, but work.
From order receipt to customer service: Where AI agents are already in use
An AI chatbot in mechanical engineering answers questions. An AI agent completes tasks. The difference is fundamental. And it is evident along the entire value chain, from order receipt to maintenance at the customer’s site.
Requirement specification analysis and offer processes: At the beginning there is the order. Automotive suppliers like Schaeffler receive thousands of requirement documents from their OEM customers every year. Schaeffler uses generative AI in mechanical engineering to automatically process system requirements – a reduction of 60 percent with over 100,000 hours of annual effort. The principle can be transferred to any ordering and quotation process: The agent extracts requirements, compares them with the product portfolio and marks deviations.
DIN standards and compliance: Before a product goes into production, it must conform to standards. An agent searches through relevant standards and compares them with the current product documentation. At Maschinenfabrik Reinhausen — a company with over 20 years of experience in machine learning — ISO standards such as 27001 and 55001 are already an integral part of AI-supported product development. For regulated industries, this not only saves time but also prevents expensive certification problems.
Material certificates and spare parts identification: During ongoing production, material certificates, test reports and certificates are generated for each delivery. An agent extracts the relevant values such as alloys, tolerances and test stamps and automatically compares them with the specifications. When searching for spare parts, the system uses photos or part numbers to identify the right component from catalogs with tens of thousands of items.
Freight invoice verification: Billing follows after shipping. An agent compares incoming logistics invoices with delivery notes, framework agreements and tariff tables. With thousands of items per month, this reduces manual checking effort by up to 80 percent. According to practical reports, the investment pays off in six to nine months.
AI in after sales — service technician assistance and knowledge database: This is where RAG technology (Retrieval-Augmented Generation) comes into play. The technician stands in front of the machine, enters an error code, and the AI assistant searches operating instructions, maintenance manuals and spare parts catalogs in seconds. The result: a step-by-step diagnosis with references to sources. Up to 40 percent fewer on-site service calls because more cases can be solved over the phone.
Two examples show how great the need is in after sales: Liebherr manages over 50,000 technical documentation topics in the mining area alone, spread across a global service network. Deutz relies on an interactive 3D service manual that automatically adjusts the level of detail to the technician’s experience. Because the knowledge must remain in the company, even if the expert leaves.
Customer complaint management and first level support: At the end of the chain are complaints and support requests. An agent classifies incoming complaints, forwards them to the correct specialist department and answers standard cases automatically from the knowledge database. First level support in mechanical engineering – often a bottleneck today due to a lack of skilled workers – can thus be largely automated. For manufacturers with a global service network such as Deutz, where over 200 technicians work around 1,600 hours per year on site at the customer in Germany alone, this significantly reduces the processing time per process.
Controlling agent: The numbers run parallel to this. Monthly financial statements, cost center reports, variance analyses. Tasks that cost controllers hours because they collect data from SAP, Excel and emails. An AI agent automatically consolidates and delivers prepared reports. Schaeffler shows where this is leading: With ML-supported forecasting, forecast accuracy in the spare parts business increased from 63 to 79 percent.
The same principle applies everywhere: technology is not the bottleneck. The bottleneck is the data.
Voices
Phillip Pham, Managing Director of Pexon Consultinga service provider specializing in AI consulting for mechanical engineers and manufacturing with 80 employees: “One of our customers had 14,000 technical PDFs on a network drive – without keywording, without structure. The AI was implemented in three weeks. The document preparation took three months. Anyone who wants to introduce knowledge management with AI will not fail because of the technology, but because of the data quality.”
Jochen Köckler, CEO Deutsche Messe AG, told the Hannover Messe press office: “Anyone who boldly invests in AI, automation and digital systems today creates the basis for leaps in efficiency, resilience and sustainable competitiveness.”
Fraunhofer IFF, KI_eeper research project on knowledge transfer: “Baby boomers who will retire in the coming years are causing companies to lose a large part of their experiential knowledge – especially the implicit knowledge in manual production activities, which is only anchored in the employees themselves.”
What this means for IT managers
The barrier to entry is lower than many expect. A pilot project based on Azure Search and Azure AI services can be implemented in eight to twelve weeks – including document preparation and integration into existing work environments. The cost of a functional prototype is around 10,000 euros.
According to Google Cloud, 78 percent of manufacturing decision-makers already see measurable benefits from AI applications. The calculation is simple: If 100 clerks each save 30 minutes of manual document work per day, this results in an annual saving of around 350,000 euros at an internal hourly rate of 40 euros.
However, the question is no longer whether AI agents will come into production. According to Bitkom, 42 percent of German industrial companies are already using AI in production – almost twice as many as in the previous year. According to the VDMA, 53 percent of mechanical engineers expect an increase in sales through AI within three years.
In the end, it is not the technology that determines the success of AI in manufacturing. But the question is whether companies transfer their empirical knowledge into an AI-supported knowledge database in good time before the experienced employees retire.
Anyone who wants to know how one AI agent in after sales Pexon Consulting has one that looks concrete Mechanical engineering case study — including architecture, schedule and budget.