AI consulting: useful, transparent and on your own premises

I help you find out where AI genuinely saves time in your business and, if you wish, implement solutions with local models where your data never leaves your network.

Book an appointment Open the AI lab

Cloud AI services are powerful, but many businesses rightly hesitate to send contracts, customer data or internal documents to external providers. Local language models are now good enough to handle many everyday tasks reliably – on a computer in your office. I advise you honestly on when local makes sense and when a cloud service remains the better choice.

What I do for you

AI potential analysis
We collect recurring tasks and assess where AI realistically helps – and where it does not.
Local language models
Selection, setup and hardening of open-source models on your own hardware or private server.
Document AI
Ask questions of your own documents, pre-sort emails, read receipts – with access control and traceable sources.
Integration into existing workflows
Connection to file storage, email or business software so that AI helps where your team already works.
AI policy for your team
Clear rules on which data may go into which tools – written in plain language and agreed with the team.
Pilot and evaluation
A manageable pilot with measurable criteria before you invest in larger solutions.

Especially suitable if …

your team spends a lot of time searching, summarising or retyping
you process confidential data and therefore avoid cloud AI
staff already use AI tools without clear rules
you want to try AI without committing to one provider

What you have at the end

An evaluated list of concrete AI use cases
A running, secured solution or an evaluated pilot
An AI policy your team understands

AI lab

Two small tools that show what local AI means. Everything runs in your browser; no data is transmitted.

Where does your data go?

Choose an operating model and watch the path a document takes.

What changes

Place of processing: Local AI: On hardware in your office or on your private server
Control over data: Local AI: Entirely yours
Costs: Local AI: One-off hardware, then mainly electricity and maintenance
Model quality: Local AI: Very good for many everyday tasks, limited by hardware
Internet dependency: Local AI: Also works offline

Both approaches have their place. In consulting we choose the right model for each use case – often a combination makes sense.

What hardware does a local model need?

Choose model size and precision. The estimate shows the approximate memory required.

Estimated GPU or system memory

≈ 4.8 GB

Typical hardware class: Modern laptop with 16 GB RAM or a computer with an Apple silicon chip
Suitable for: Email drafts, summaries, simple questions about documents

Rule of thumb: parameters × bits ÷ 8 plus around 20% for context and runtime. Actual requirements depend on the model, context length and number of simultaneous users.

Common questions

Are local models as good as large cloud services?

For many everyday tasks such as summarising, classifying or querying documents they are perfectly adequate. For very complex tasks, large cloud models are often superior. We check this for the specific use case.

What hardware do we need?

That depends on model size and usage. The AI lab on this page gives a first orientation; the exact sizing is part of the consultation.

What about data protection and the EU AI Act?

Data protection and regulatory requirements are part of every plan. For a binding legal assessment I work with specialised legal advisers where needed.

Let's spend 15 minutes talking about your IT.

No obligation and free of charge. You describe your situation, and I'll tell you honestly where I would start – or whether you need help at all.

Book online Send a message

What I do for you

AI potential analysis

Local language models

Document AI

Integration into existing workflows

AI policy for your team

Pilot and evaluation