AI infrastructure knowledge base.
Our AI infrastructure knowledge base brings together four formats: articles on current questions, a German-language AI glossary for technical terms, implementation guides for concrete projects, and comparisons between approaches and tools. All of it is written for decision-makers in the Mittelstand, with no assumed technical background and a clear tie to real business questions.
What's on the shelf
Articles take on questions that keep coming up in our projects and that don't fit into two sentences. They aren't tied to an offering. When the facts change, we change the article instead of writing a new one. On the shelf: How long does an AI implementation take?
The glossary tells you what a term means when you meet it in a proposal or a tender. One entry, one thing, without a chain of references to work through first. On the shelf: What does private AI infrastructure mean?
A guide walks alongside a project that has already been decided: what happens in which order, where it usually stalls, what should be settled beforehand. Written for whoever has to prepare the decision. On the shelf: From pilot project into operations
A comparison puts two routes side by side and says from what point each one holds up. Usually that comes down to a number. We don't recommend either option outright, because the answer depends on volume, on your data and on your time horizon. On the shelf: Own GPUs vs token pricing
Knowledge Articles & Guides
In-depth analyses, practical guides, definitions and cost calculations for enterprise AI.
Intralogistics automation: what a sequencing process delivered in six weeks teaches
Intralogistics automation pays off particularly where manual, recurring transport and sequencing steps set the pace. A documented case from VW's Gläserne Manufaktur shows, according to the company, that such a process can be fully automated with transport robots, a cobot and adapted load carriers, implemented in six weeks.
AI use cases in manufacturing: the seven fields at a glance
AI use cases in manufacturing fall into seven fields: predictive maintenance, quality inspection, planning and supply chain, energy and process, knowledge and assistance, engineering and back office, and agentic systems. They differ widely in maturity, typical impact and data requirements. Your own data situation, not the technology, should therefore decide where you start.
AI for mid-sized companies: status, barriers and a plan for the first 90 days
AI for mid-sized companies is more widespread than often assumed, but unevenly distributed: use rises with company size and even more with in-house research and development. It is held back mainly by data protection, the skills shortage and lack of time, hardly by technology. A narrow start with a short payback, prepared in 90 days, makes sense.
Choosing an AI service provider: 14 questions to ask before you sign
When choosing an AI service provider, focus less on models and demos and more on three interfaces: where the system runs, whether the implementation actually reaches production at the agreed price, and who then operates the system with written commitments. Add to that a defined exit, references with figures and a clear data classification.
ZIM funding freeze alternatives: which funding sources remain for AI projects
As of August 2026, the most important ZIM funding freeze alternatives for AI projects are KMU-innovativ, with its next deadline on 15 October 2026, regional programmes such as Digitalbonus Bayern and, in Switzerland, Innosuisse. The former standard federal programmes have expired. Plan your project so that it pays for itself from its own business case.
Why AI projects fail: an implementation problem, not a technology problem
Why AI projects fail can be answered clearly for industry: rarely because of the technology, mostly because of the implementation. What is missing is data access, a baseline before the start, clear usage rules and operational ownership after the pilot. According to Deloitte 2026, only one AI use case in five in manufacturing is scaled across the company.
Open model weights explained: what they are and why they matter
Model weights are the numerical values a language model learned during training, and they determine how it calculates an answer from an input. Open model weights means that the developer publishes these values as a file, so you can run the model on your own hardware, subject to the terms of its licence.
Digital sovereignty for AI: how to make it verifiable
Digital sovereignty for AI only exists once you can prove where your models run, which data leaves your premises, who used what and how you can leave the provider again. It becomes verifiable through four criteria: location of processing, data flow, complete logging and the replaceability of the model.
How long does an AI implementation take?
In practice, an AI implementation rarely takes longer than ten weeks to reach productive use. After a one-week assessment, a pilot is running internally within 72 hours; the rollout that follows takes 4 to 10 weeks depending on scope and system landscape.
Editor's Choice
Run Models Bigger Than Any of Your GPUs: SilentSwarm Is Now in Open Beta
SilentSwarm splits a large language model into pipeline stages and spreads them across the mismatched GPUs you already own, even behind NAT and firewalls. Train, fine-tune and chat as if it were one machine.
The Local AI Operating System: GDPR-Compliant LLMs on Your Own Server Infrastructure
vLLM, LiteLLM, Langfuse and OpenWebUI on your own GPU: the most direct path to data sovereignty, GDPR compliance and predictable AI costs.
The EU AI Act – What will happen on 2 August 2026?
2 August 2026 is the first hard deadline under the EU AI Act. From that date onwards, certain AI systems will be banned in the European Union – not regulated, not restricted, but illegal. Anyone who continues to operate them after that date risks fines of up to 35 million euros or 7 per cent of their company’s global annual turnover.
Private AI Agent Systems with Orchestrated Specialist Models
How private multi-agent systems combine small specialist models, MCP, A2A and cost-aware routing into secure enterprise AI architectures.
Time Series Forecasting with PatchTST: Why Classical Models Are No Longer Enough
Why modern time series forecasting needs more than ARIMA and Prophet: patch-based Transformers, rigorous benchmarks, and a platform that connects ETL, training, monitoring, and forecast APIs.
Why we write this
We write these pages because we answer the same questions in every first call: what it costs, how long it takes, what happens after go-live, and what all of that depends on. They're good questions. They don't belong in a sales conversation where you can only answer them under time pressure.
Which is why the answer sits here instead, readable without an appointment. If you finish reading and conclude you don't need this, or don't need it from us, the page has done its job. The shelf grows slowly: we write once a project has taught us something that holds true beyond that project.
When reading isn't enough
At some point the general answer stops helping, because the next question depends on your own organisation: your data, your department, your volume. There are three routes from here. The tax advisors industry page shows how these topics look in one specific field. The cost calculator gives you a first ballpark to work with internally. And if you have a question nothing here fits, write to us via the contact page. We answer even when it doesn't turn into a project.