I support manufacturers in the Siegerland and Altenkirchen district with AI, from the first question through a pilot on your hardware to the rollout. The first consultation is free.
A supplier from the region had AI in the company long ago, consciously and unconsciously. Shadow IT has long existed in companies; employees improvise to bridge gaps. Sales had quotations drafted, engineering tried something else, and in quality assurance a folder of test protocols sat there that no one could search. Bans neither solve the problems nor prevent uncontrolled data outflow.
The question was never whether AI helps. The question was where it starts and which data leaves the company in the process. That is where we begin, process first, technology second. Instead of one large project, we first clarify where the lever is in your company, which data has to stay local and what a first step looks like that fits the current budget.
This area takes you step by step. Is AI worthwhile for us at all? What benefit does it bring in your areas? Which data may leave the company, what do you have to watch out for legally, which models and costs are realistic? Start with the topic that concerns you right now, or read in order.
They do not fail on the technology. The sore point is the question of which process is actually being looked at. That is exactly where I start. Before we talk about tools, models or hardware, we clarify where the lever is in your company and whether a project pays off at all.
Process before technology. This keeps the first step small, below the investment threshold, and the result arises within the current budget year instead of in the next planning round.
At a plastics company, the topic seemed to be replacing injection-moulding machines. As soon as you factored in the upstream and downstream steps, the incidental costs and the question of whether the manual rework can even still be staffed, it came down to the business. What price the end product bears in future, what margin remains and how the competition positions itself. In the end there was a different decision than the one you had started with.
That was not an AI project, but it shows the pattern. The first question is almost always cut too small. What was laborious was bringing the numbers together. That is exactly what there are tools for today that did not exist back then. They do not make the decision. They put the basis for it onto the table in days instead of weeks.
The figures support this. Among the most common obstacles to using AI are legal uncertainties at 53 percent, a lack of technical know-how at 53 percent and a lack of personnel resources at 51 percent. Not a single one of the most frequently named hurdles is a question of model quality.
Source: Bitkom, Artificial Intelligence in Germany, study report 2026, survey of 604 companies with 20 or more employees.
Before we talk about tools, I go through six questions. They decide whether a project holds up, and often shift what is even being talked about.
Often the visible task is only the repair of another problem. If the search for test protocols takes two days, it is less about speed. The real question is why searching is necessary at all.
What for? In two minutes you see where your company stands and what the sensible next step is. Answer the eight questions and the result, including a recommendation, appears automatically.
Answer the questions and your classification with a concrete recommendation appears here.
Question five is the most revealing. In most companies the answer is yes, without management knowing.
The quick check tells you where your company stands. If you have a specific process in mind and want to know whether it will carry an AI agent, the whitepaper takes it further: thirteen questions and five levels of build-out, free to download.
The entry is deliberately cut so that it triggers no investment application. You decide after each stage whether the next is worthwhile.
One hour plus a written assessment on two pages. It names the viable use case, its risk class and the next step.
A moderated half-day plus a pilot on existing hardware with your real documents. No purchase, no server procurement.
From the pilot into regular operation, with hardware, permissions, connecting data sources, training and evidence.
On your actual use cases and the data rule from stage 2. As a separate module or embedded.
We clarify terms and funding options in the first consultation.
The economic structure here is unusually homogeneous, which makes the work easier. In the district of Siegen-Wittgenstein, metal production and processing, the manufacture of metal products and mechanical engineering form the core of industry, with many automotive suppliers among them. There, the producing sector accounts for around 43 percent of gross value added and almost 39 percent of employees subject to social-insurance contributions (IHK Siegen, Überblick in Zahlen 2025). In the district of Altenkirchen, mechanical engineering as well as metal and plastics processing form the focus; the manufacture of metal products accounts for the largest share of industrial employees at around 29 percent (Industriekompass Rheinland-Pfalz 2024, data year 2023).
Five industries thus cover the largest part of the use cases. Experience from one company can often be transferred to the next.
I have worked in industrial companies for around thirty years, continuously in the environment of metal and plastics, at the interface between marketing, sales, product development and engineering. Anyone who over the years has to prepare technical matters so that they become understandable for customers cannot avoid a solid understanding of the subject.
I am not a production planner. When it comes to a specific plant, I go into production and have the process explained to me by those who run it every day. With AI models, operating modes and data architectures I am at home professionally. I just do not start there.
What companies gain. You get transferable experience from a homogeneous industry structure. On-site appointments instead of remote diagnosis. And an entry that fits your current budget without an investment application. Further cases are set out under Stories from practice.
The first consultation is free; you receive a clear inventory and assessment. Only the pilot and the later rollout are chargeable, and you decide after each stage whether the next is worthwhile.
Not necessarily. Many AI use cases in the mid-market can be implemented locally or in a hybrid setup. Data does not have to go into the cloud, or not entirely. What is permissible for which data class we clarify in conversation.
A pilot needs adequately dimensioned hardware to run the right language model performantly for the requirements. First results often show within the first week. After that we talk about the necessary investments, once it gets concrete.
I work supra-regionally throughout the German-speaking area, on site and remotely. A particular focus lies on manufacturers in the Siegerland and the Altenkirchen district, such as metal and plastics processing, mechanical engineering and automotive suppliers. The homogeneous industry structure there makes experience transferable.
Your AI project begins with the right considerations and questions. In a free first consultation of about an hour, we clarify whether there is a lever for you and what a first step looks like that fits your current budget.