Skip to content
PACKAGES · CLOUD-FREE AI
AI Workroom

Cloud-Free AI

We set up AI on your own devices for data that should not leave the premises.

Add-on to the AI Workroom package.

Starting point

When this add-on is for you

Your AI Workroom is running. The tools are set up, your team works with them, and the operations manual states which data may go to the cloud.

However, part of your work cannot be handled through the AI Workroom. Quite possibly, that is exactly where there are processes in which AI support could also save a lot of time: analyses of your own figures, your contracts, costings, HR documents, drafts. In short: your own business data that you don’t want to leave your premises.

So that you can use AI for this data too, we set up a second system for you: an AI environment on your devices that does its work without the cloud. Whatever you put into this environment is processed on the device and does not leave it via the environment. The device itself remains your usual workstation, with network, email and your programs.

Our solution

We set up an AI environment on two of your devices that works without the cloud. Together with you, we define which tasks belong there and which remain in the cloud environment of your AI Workroom.

First, we check by video call what your devices can handle; this also gives a minimum requirement for later purchases. We propose the local model to you in writing; you decide. Then we set up the environment, create templates for the agreed tasks and record the set-up state with a date.

In the training, management and team work on a real, confidential task from your business on their own device. In the 90 days that follow, we make adjustments where things turn out differently in day-to-day work than expected.

Robot in a station holding a network cable next to a socket labelled Internet
OUR ADD-ON TO THE STAR CHART PROGRAMME: YOUR ROUTE TO LOCAL AI ON YOUR OWN DEVICES.
The path of the files

The file goes in, the result comes out – in-house

CONFIDENTIAL DATA · NO WAY FOR ITLOCAL PROCESSING · IN-HOUSE
Local environment ready.
Ready for the digital future.
01 · Suitability check per deviceWhat your devices can handle, recorded in writing per device.
02 · Assignment by type of dataWhich task runs locally and which stays in the cloud.
03 · Choice of the local modelCandidates tested against your cases, recommendation in writing.
04 · Set-up on two devicesTuned, tested, frozen and handed over with documentation.
05 · Templates for up to four tasksOne template per task, plus two quick guides.
06 · Training on a real taskA confidential task from your business, processed locally.
07 · Follow-up sessionsThree sessions after 30, 60 and 90 days from sign-off.

The facts

Finding 01 · Adoption
Estimate

Which reason do small companies that considered AI but did not introduce it cite most often?

73 %
CITED LACK OF KNOWLEDGE · SURVEY 2025
73 % CITED LACK OF KNOWLEDGEYour estimate
64 % CITED UNCERTAINTY ABOUT THE LEGAL CONSEQUENCESYour estimate
60 % CITED DATA PROTECTION CONCERNSYour estimate
46 % CITED INCOMPATIBILITY WITH EXISTING SYSTEMSYour estimate
33 % CITED COSTS TOO HIGHYour estimate

Money stands in the way less often than knowledge.

Lack of knowledge is the most common reason: 73 per cent cited it, costs too high only 33 per cent.

Surveyed in 2025 among companies with 10 to 49 employees that do not use AI but have considered using it. Multiple answers were possible. Second most often cited: uncertainty about the legal consequences (64 per cent).

Measured value hidden
Estimate first, then see.

Choose an answer above to reveal the measured value.

Source: Statistisches Bundesamt, table “Gründe gegen die Nutzung von Technologien der künstlichen Intelligenz nach Beschäftigtengrößenklassen”, survey 2025, column 10 to 49 employees
Method and limitationsPopulation: companies with 10 to 49 employees in Germany that do not use AI but have considered using it · multiple answers possible · sample size of this subgroup not published; net sample of the survey 20,000 companies, participation voluntary (Eurostat, national reference metadata Germany “ICT usage in enterprises (isoc_e)”, item 13.3.3.1.1)
Finding 02 · Distinction

An AI account alone is not AI management.

An account can be set up in a few minutes. Since December 2023, an international standard has described what sets orderly AI use apart from that: ISO/IEC 42001 describes what makes up an artificial intelligence management system – established, implemented, maintained and continually improved. The standard explicitly applies to organisations of any size. An account alone does none of this.

COUNTS AS AI MANAGEMENT
  • Rules defined and written down
  • Responsibility assigned by role
  • A process that is reviewed and adjusted

The business has defined who may work with what, which data goes in and who is responsible for the result – written down and repeatable.

DOES NOT COUNT
  • Access set up, no defined rules
  • Verbal agreement, no responsibility by role
  • Processes not written down, only in individual people’s heads

Access to an AI system with no defined scope, limit or responsibility.

Distinction based on ISO/IEC 42001:2023, artificial intelligence management system, published on 18 December 2023 – a standard, not a survey. Application is voluntary.
Context

What holds businesses back is less often price or technology than a lack of knowledge and a lack of clarity about what applies. Both can be put in place: through instruction and through rules that are written down. What that means for your business depends on which tasks are to run through the system, which data passes through it and who is responsible for the results. We clarify that together with you in the introductory call.

Method and limitationsAbout the figures: the official survey covers companies with ten or more employees. For smaller companies, Germany reports no figures on AI use or on the reasons against it; there are no reliable figures for them. In addition, the shares cited refer only to companies that have already considered using AI, and multiple answers were possible.

What to expect in this add-on

PROCESS · SEVEN STEPS

We check by video call what your devices can handle and record the result in writing for each device – including a minimum requirement for later purchases. We do this in advance, so that before the first session at your premises it is clear what we can build on.

Together with you, we define which task will run through which system in future, and tell you which tasks are suitable for your AI systems. These are the policies that remain after the handover, even when a new task is added or someone new joins the team. We write them into your operations manual, where the guidelines from the main package already are.

We work with a selection of open models that are commercially usable. Which one suits you depends on four things: what your devices can handle technically, which tasks are to run through it, what matters to you and whether you prefer a particular model. We test the candidates against your cases and present a recommendation to you. You can then decide.

Set up, tuned to your tasks, tested against the agreed test cases. The state is frozen and handed over with documentation.

One template per task that your team can use straight away, plus two quick guides for the steps that are the same every time.

In the training, we work on a real, confidential task from your business: locally on your device, with your data. First individually with your management, then as a block with your team.

Three sessions after 30, 60 and 90 days from sign-off, by video call, already included in the fixed price. We check whether your system holds up in day-to-day work and make adjustments where things turn out differently than expected.

DOCUMENTS AND FORMATS

For your types of task, it is set which system they belong in and which tasks are unsuitable for AI. Anyone who is unsure can look it up instead of having to decide for themselves. We write this where the guidelines from the main package already are.

We work with a selection of open models: no licence that keeps running every month, no dependence on a provider’s plan. The state handed over is frozen, dated and documented; it stays as it is until you change it.

Is this add-on right for you?

  • For day-to-day work

    Everything comes together with you: you analyse figures, write drafts, work with contracts. For the confidential ones among these tasks, an environment that works without the cloud is then ready.

  • For the in-house IT contact

    You get asked when something is unclear. Day-to-day operation is in your hands: we go through what that involves in the training and write it into your operations manual.

  • For owners and managers

    You set the course and decide which data may leave the premises. After the handover, it is set for each type of task which system it runs through.

The add-on at a glance

Price
from €5,240.00*
Package €5,240.00 · Add-ons €0.00
Price and payment
Complete package€5,240
Paid in instalments as the work progressesin the quote
Add-ons
In the quote
Terms and notes
Workstation with a chat window: internet offline, AI system ready to go, the network socket next to it

Common questions

Whatever you put into the local environment is processed on the device and does not leave it via the environment. However, some set-ups can undermine local processing: for example a working folder that is synchronised to cloud storage, a backup that writes your files to another location, or another application that passes the same data on externally. We point out these cases to you and write them into your operations manual. The device itself remains your normal workstation, with network, email and your programs.

We clarify that in the suitability check in advance, by video call: we look at your devices and record in writing for each device what it can do. If your devices are sufficient, there are no further costs for you. If they aren’t, we tell you in the suitability check what a suitable device needs to be able to do. We record the minimum requirements in your operations manual; they also apply to later purchases.

Yes. The prices on our website are starting prices for the standard case. Based on your request, we first review what is already running in your business and what the project needs. You then receive a written quote with the scope, a fixed price and a completion date. We stick to it; there are no additional charges later. Payment is made in instalments, each after you have signed off a stage.

The hardware. The add-on is designed for two devices with a local set-up and up to four task templates. For additional devices or more templates, we present an individual quote to you before you decide. The legal classification of your data remains with you; we recommend legal advice for this.

We work with a selection of open models that are commercially usable: among others, models from Mistral (e.g. Mistral 4 Small) and Google (e.g. Gemma 4 26B A4B QAT), Qwen models, open models from OpenAI such as GPT OSS and, where a task requires it, specialised models. The list is not exhaustive and is continually extended as development progresses. Which one suits you depends on four things: what your devices can handle technically, which tasks are to run through it, what matters to you and whether you prefer a particular model. We test the candidates against your cases and present a recommendation to you. You then decide. The choice also includes the technical specifications: whether you prefer a so-called “dense model” or an “MoE model” (Mixture of Experts), how many parameters it may have and in which quantisation it should run. This also depends on the capacity on your hardware that may be reserved for the models. You don’t need to remember any of this: we explain the background to you, set everything up for you and document it in your operations manual.

You do. A note on this: we only set things up and train your team; which data you then enter and which legal conditions apply to it is then up to you. Please also note any legal obligations you have to observe as the operator of the systems. If needed, we recommend a brief consultation with your legal adviser before the project starts.

Yes. You receive a frozen, dated and documented state. It stays as it is until you change it. If you later want to bring it up to a newer model version, you can commission us separately to do so. We check whether a newer version gives you advantages, install it, run your test cases and freeze the new version again, documented.

We come to your business twice: once for the set-up on your devices and once for the training. At no extra charge within 100 kilometres of our base in Trier: all of Luxembourg, from the Oesling to the Minett, and on the German side, for example, Schweich, Föhren, Wittlich, Saarburg, Bernkastel-Kues, Cochem, Koblenz and Kaiserslautern. Everything else, the suitability check, coordination and the three follow-up sessions, takes place by video call, throughout Luxembourg and Germany at no surcharge; if you are outside the 100 kilometres and would like on-site sessions, we put a figure on the travel costs in the quote.

The AI Workroom package should have been completed, immediately before or earlier. The Cloud-Free AI add-on builds on it: we write the policies where the guidelines from the main package already are. If your business already meets these requirements, or if you explicitly want to use only AI without the cloud, this add-on can also be commissioned as a standalone package. In that case, please contact us directly; we review your request and present an individual quote to you.

Discuss your mission

Describe your starting point. We’ll reply within two working days.

Our form needs JavaScript.

JavaScript is turned off in your browser. Please send your enquiry directly to kontakt@0674nord.de, or turn on JavaScript and reload the page to use our form.

Please tell us your name.

That doesn’t look like an email address.

Please briefly describe your enquiry.Please describe your enquiry in a little more detail.Please shorten your enquiry to no more than 4,000 characters.

Your details are transmitted in encrypted form. We don’t need sensitive data such as login or bank details for the introductory call; please leave them out here.

Product sheet (PDF, in German)