ARTIFICIAL INTELLIGENCE

AI where it takes
work off your hands.

Not every problem needs a language model. But some work one does in seconds: searching contracts and manuals, pre-answering the same requests over and over, summarising minutes, reading documents. We look at where your time gets lost and tell you whether AI helps there or whether simple automation is enough.

Book a consultation Local or cloud?
WHERE AI HELPS

Recurring work
nobody enjoys

  • Finding knowledge

    Your team asks in its own words and gets the answer from manuals, contracts and minutes, with a reference to the source. How it works

  • Pre-answering requests

    Recurring questions by email or ticket get a suggested reply. A person checks it and sends it.

  • Reading documents

    Details from letters, forms and delivery notes end up where they belong instead of being typed in.

  • Summarising

    Long threads, meetings and reports cut down to what matters.

And sometimes the answer is not AI If your supplier sends an e-invoice, every detail is already neatly in the file. Then we read it directly, without a language model. That is faster, cheaper and misreads nothing. We always take the simplest route that works reliably.
YOUR KNOWLEDGE · RAG

Answers from your documents,
not from the internet

A language model does not know your company. It knows neither your manuals nor your contracts or your prices. Ask it anyway and it guesses. The remedy is a knowledge base the model searches first for every question. The technical term is retrieval-augmented generation, RAG for short.

1 · Ingest
Your documents are split into sections and made searchable: manuals, contracts, minutes, wiki, tickets, folders on the server.
2 · Retrieve
For every question the system picks out the matching passages, even if they use different words.
3 · Answer
The language model writes the answer from exactly these passages and names the source.
  • No training needed

    The model is not trained on your data. New documents can be found immediately, deleted ones disappear again.

  • Every answer with its source

    Below the answer you see which document it came from. One click and you see the original.

  • Permissions stay

    If you wish, everyone finds only what they may open in the original. The personnel file stays with HR.

  • Runs locally too

    Knowledge base and language model can run entirely locally at your site or in a German data centre.

What RAG cannot do Made-up answers become much rarer, but they are not ruled out. That is why the source is always shown. And the answers are only as good as the documents: outdated documents give outdated answers. We help with the clean-up.
YOUR DATA

Your data stays
where you want it

We run language models locally at your site or in a German data centre. Your documents then do not end up with any third-party provider. If a cloud model is clearly better for a task, we tell you so, and also what happens to your data. The decision is yours.

Often it is a mix: confidential material stays with the local model, public texts may also be seen by a cloud model. If we operate the model for you, you sign a data processing agreement with us.

AspectLocal modelCloud model
Where the data is processedLocally at your site or in a German data centreAt the provider, depending on the contract also outside the EU
QualityGood for searching, summarising and sortingStrong even on long and difficult tasks
CostServer or hardware, no cost per requestNo hardware of your own, cost per use
Who decidesYou decide when a model is updatedThe provider changes models and prices
WHAT WE DO

From the first conversation
to day-to-day operation

Advice

We find the processes where AI brings the most, and the ones where it has no business.

Initial consultation free

Integration

Into your software, intranet or shop, via interfaces and with access only to the data the model needs.

Scope by agreement

Your own AI servers

Local language models and knowledge base (RAG) on your own hardware, reachable only by your applications and your team.

Local or German data centre

Operation

We keep models, servers and interfaces up to date and secure, and notice when something goes wrong.

Ongoing

Training

Anyone using AI has to make sure their team can handle it; the EU AI Act has required this since February 2025. We show your people how to do it safely.

On site or remote
Already in use We already run local language models with their own knowledge base in production for customers.
FREQUENT QUESTIONS

Answered briefly

Do we need our own hardware?

Not necessarily. A local language model runs on a server with a graphics card locally at your site, or on a server in a German data centre that we operate for you. What you need depends on the model and on how many people use it. We work that out in the initial consultation.

Is a local model as good as ChatGPT or Claude?

For many tasks it is good enough: searching your own documents, summarising texts, sorting and pre-answering requests. For long and difficult texts the large cloud models are usually better. We tell you up front what to expect.

Do we have to train the model on our data?

No. With a knowledge base (RAG) the system looks up the matching passages in your documents for every question and passes only those to the model. The model itself stays unchanged. New or changed documents are therefore taken into account immediately.

What happens to our data?

With a local model it stays at your site or in our German data centre and is not used for training. With a cloud model we show you, before you decide, what the provider does with the data under its contract and where it processes it.

What does the EU AI Act require of us?

Since February 2025, companies that use AI must make sure their staff have sufficient AI literacy (Art. 4 AI Act). Which further obligations apply depends on the use case. You clarify the legal assessment with your legal advisers; we provide the technology and the training.

Do you also add AI to software you did not build?

Yes, as long as the software has an interface or can be extended. That includes shops, intranets and ticket systems.

What does it cost to get started?

The initial consultation is free. Afterwards you get a plan with effort and cost before we build anything.