Local AI,
explained without theatre.
Practical research for professionals deciding what to run locally, how to evaluate it and where the operational boundary belongs.
Start with the decision you need to make.
Each article states its scope, sources and update date. Claims are written to be inspected, not merely repeated.
What does it mean to own your AI?
A practical definition of AI ownership across hardware, data, models, operations and exit rights.
Local AI vs cloud AI for German businesses
A workload-by-workload comparison of local and hosted AI for German SMEs and professional offices.
Local LLM hardware requirements: RAM, models and quantisation
A sober sizing guide for running 8B, 14B, 32B and 70B-class models on local hardware.
Private AI for law firms and tax advisers
Where local AI can help professional offices, where human review remains mandatory, and how to set a defensible boundary.
Using local AI with confidential business documents
A controlled document workflow from approved folder to local index, answer, review, backup and deletion.
Open-weight models for German and European documents
How to select and test multilingual open-weight models for German, European and document-heavy work.
From research to a working policy.
Role guides and reproducible tests turn the broader argument into concrete acceptance criteria.
Need a system for a real workload?
Describe the documents, users and response expectations. We will map them to a Core or Pro appliance and model class before configuration.