Made for Swiss self-employed & SMEs
You do not need a subscription to use AI. A perfectly capable model can run on the laptop you already own — quietly, on your own hardware, with the data never leaving the room. For a Swiss notary, fiduciary or doctor, that is not a gimmick. It is the difference between sending a client file to a distant server and keeping it on the desk in front of you.
Local AI sounds technical, but the everyday version is simple. You install one small application, it downloads a model to your computer, and from then on you can summarise, draft and translate — offline, in private, without a monthly bill. This guide walks you through what you need, what works well, and where to start.
What local AI actually means
Most AI you already know — ChatGPT, Claude, Copilot — runs on servers you do not control. You send your text to a data centre, and an answer comes back. Local AI is the opposite: the model itself runs on your computer. Your document, your question and the answer all stay on your own machine.
A model is just a file — a few gigabytes of learned patterns. Once it sits on your disk, it behaves like any other program: you open it, use it, close it. Nothing is sent anywhere.
Is my laptop strong enough?
The honest answer is: for the small jobs, probably yes. The newer models are surprisingly light.
- A Mac with an M1 chip or newer (from 2020) runs small models comfortably.
- A Windows or Linux PC with 16 GB of memory and a recent graphics card handles the smaller models well.
- A modest machine can still run a compact model for short summaries and drafts — it is just a little slower.
Start with a small model. If it feels too slow or too weak, you can always move up — but most first users are surprised how much a small model can already do.
The tools to start with
There are two friendly starting points, both free:
Ollama
Ollama is a small command you install, then type one line to download and run a model. It stays in the background and answers from a terminal window.
LM Studio
LM Studio is an app with a normal window, a search field for models and a chat-style interface — it feels familiar from the first minute.
Both run the same kind of models. Pick whichever feels more comfortable; you can switch later.
What local AI is good at — and what it is not
Be realistic and you will not be disappointed:
- Good at: summarising documents, drafting emails and letters, proofreading, turning notes into clean text, translating between German, French and English.
- Weaker at: the hardest reasoning, very long documents, or anything where a small mistake matters and no human checks the result.
A small local model is a diligent assistant, not a specialist. For routine office work — exactly the kind that eats your week — it is more than enough. For the final check, you are still the one in charge.
A calm first exercise: summarise a document
Here is a good first test, using a letter or a short contract already on your desk. Do not use a real client file yet — start with something harmless, like an insurance letter or an old meeting note.
Summarise this letter in five plain sentences. List the deadlines, the amounts and anything I need to do. Keep it in the same language as the letter.
Read the result. If it is clear and correct, you have just learned what the tool is for. If something is off, you have also learned something useful: where you still need to look yourself.
When local beats cloud, and when it does not
Choose local when:
- the document contains professional secrecy, patient data or client details you would not email to a stranger;
- you want privacy without reading a vendor's terms;
- you want a fixed tool with no monthly fee.
Stay with a cloud tool when:
- you need the strongest model for a hard, non-sensitive task;
- you need a feature local tools do not yet offer;
- the convenience of your team's current setup matters more.
Many practices settle on a calm middle path: local for the sensitive routine work, a cloud tool for the rest, and clear rules about which is which.
Data rules worth writing down
Local AI keeps your data on your machine, but the same habits still apply:
- Keep the human in front — every draft is read by a person before it goes out.
- Back up what matters — a local model will not save your files for you.
- Stay current — update the app when it asks, so you run a maintained version.
One page, three lines, done. The point of local AI is less worry, not new paperwork.
The outcome
You can have a useful, private assistant running on the laptop you already own within an hour — no subscription, no data centre, no fuss. Start with one small model and one real task. Let it summarise, draft and tidy the routine work, keep the sensitive files at home, and you have taken the calmest first step into AI there is.
