Give your company an AI brain that you own.
We turn how your team works into plain files and AI skills. Your data stays where you decide: Swiss-hosted, zero-retention, or fully local on your own hardware. One team, from the first call to ongoing care.
of Swiss SMEs build AI into their processes on purpose. 39% more are testing it.
AXA-Sotomo SME labour market study 2026, AXA press release, 15 Sep 2026 (read 15 Oct 2026).stay responsible for your data. Swiss data protection law applies to AI tools, whoever runs them.
FDPIC (Swiss data protection authority), notice of 9 Nov 2023, as reported by Netzwoche on 10 Nov 2023. Not legal advice.A Swiss open AI model exists. Apertus, built by EPFL, ETH Zurich and CSCS, is available in Infomaniak's Swiss assistant Euria.
EPFL; Infomaniak FAQ (read 15 Oct 2026).One place where your company's knowledge lives. In files you own.
A second brain for your team: a folder of plain Markdown files, opened in Obsidian. Your people read it. Your AI reads it. If a tool goes away, the files stay.
The AI stops giving generic answers. It knows your offers, your clients, your rules and how you work.
Book a free callcompany-brain/ ├─ AGENTS.md how the AI should work here ├─ Vision/ positioning, voice, goals ├─ People/ who does what ├─ Domains/ sales, delivery, support, admin ├─ Projects/ work that has an end ├─ Skills/ your methods, step by step ├─ Tools/ what is connected, and how └─ Weekly/ what changed, what is next
Plain files on your own disk. Not a subscription. Nothing to cancel.
Each task goes to the model that fits its data.
The model is rented. What makes it useful is yours. So we choose the model for each kind of data, and you can change it later.
Hosted assistants
For: Public text, drafts, ideas with no client data.
Where data goes: To the provider, under its terms.
Trade-off: Best quality and speed. Least control.
Open models, zero retention
For: Client and team data, day to day.
Where data goes: To a provider that keeps nothing and trains on nothing, in Switzerland or the EU where possible.
Trade-off: A contract and a setup to check. Often lower cost.
Local on your machine
For: Health, legal, anything that must never leave your network.
Where data goes: Nowhere. The model runs on your own hardware.
Trade-off: Needs a strong computer and upkeep. We size it, install it and look after it.
Before you sign, we name each tool and where it processes data, in writing, so your lawyer can check it. We do not give legal advice.
AI that runs on your own hardware. Nothing leaves your network.
For health, legal, finance and any data you must keep inside. We choose the model, size the machine, install it, connect it to your company brain and keep it running.
Not sure you need it? We tell you honestly. Many tasks do not.
Talk to us about local AI- 01
Choose the model
Open models such as the Swiss model Apertus, tested on your own tasks.
- 02
Size the machine
The model must fit in memory. A laptop runs small models. A shared machine on your office network can serve the whole team.
- 03
Install and connect
We set it up, link it to your company brain and your tools, and lock down access.
- 04
Keep it current
New models, updates and checks, so the setup does not go out of date.
16 GB
Small models. Drafts, simple search, learning.
32 to 64 GB
Mid-size models. Good for daily work.
128 GB and more
Large open models. A shared machine for a team.
A rough guide, not a quote. We size the machine on your own tasks and test before you buy hardware.
One team for the whole path. Take one step or all of them.
From the first conversation to ongoing care. Start small and decide with facts. Every piece we build stays yours.
Book a free call- 01
Consulting and privacy plan
Which data goes to which model, who may use it, and what we write down for your lawyer.
Written price before it starts - 02
Company brain setup
Your vault in Obsidian, the AGENTS.md file, and a guided session to fill in vision, people, domains and tools.
Written price before it starts - 03
Local and private AI setup
Model choice, hardware sizing, installation and access control. Or a Swiss-hosted option with zero retention.
Written price before it starts - 04
Skills sprint
Your team maps its processes. We write the top ones as AI skills and test them on your own examples.
Written price before it starts - 05
Connections and automation
Mail, calendar and your tools, connected with care. Repeatable steps run on their own, and a person checks what matters.
Written price before it starts - 06
Team onboarding
Your staff learn the system with their own tasks. A short weekly routine keeps the brain current.
Written price before it starts - 07
Ongoing care
Monthly upkeep, model updates, new skills, security checks and support. Optional.
Written price before it starts
From the first call to ongoing care. You see the next step before you pay for it.
Free call
You tell us what takes your team the most time. We say clearly if we can help.
Consult and plan
We agree which data may go where, and which tier each task needs. This comes before any installation.
Build
Company brain, models (hosted, Swiss-hosted or local) and the first skills.
Pilot
One process, tested on your examples. At first, a person approves each result.
Train and hand over
Team session, short guide and weekly routine. You can run it without us.
Care
Updates, new skills, checks and support, as your needs grow.
No black box. No lock-in.
The files
Your company brain is plain Markdown on your own disk. Any AI tool can read it.
The skills
Your methods, written step by step in your own folder. Not inside a vendor's product.
The choice of model
Change the model or the provider later. The files and skills stay.
The last word
The AI drafts. Your people approve, until you decide otherwise.
From our web projects. Our AI service is new.
“I am very satisfied with their services and recommend soflow to anyone who is looking for an innovative and forward-oriented web partner.”
“The speed at which our concerns were addressed and implemented exceeded our expectations.”
Where does your company's knowledge live today?
Tell us. We reply with a clear yes or no, and a time for a free 30-minute call.
Common questions
Short answers. The rest we explain on the call.
Why not just use ChatGPT?
You can, for public work. But it does not know your company, and what you paste in leaves your control. The company brain gives any model your context. The privacy plan decides what may be sent where.
Can the AI run fully on our own computers?
Yes. Local AI keeps everything inside your network. It needs enough memory for the model. We choose the model, size the machine, install it and keep it updated. Many tasks also run well on Swiss or EU hosted models with zero retention. We tell you which tier each task needs.
Why Obsidian?
Because it opens plain files. Your people get an easy app, and the AI reads the same files. If Obsidian goes away, the files stay.
We already use Notion or another tool.
Keep it. We can connect the AI to what you have, read-only at first. We do not move everything at once.
What about Swiss data protection law?
The law makes you responsible for your data, whichever tool you use. We write down where data goes and who processes it, so your lawyer can check it. We do not give legal advice.
Do we need technical skills?
No. We set it up and teach your team in plain words.
Who looks after it after launch?
You can, with a short weekly routine that we teach at handover. Or we do it with ongoing care: upkeep, model updates, new skills, security checks and support.
How much does it cost?
You get a written price for each step before it starts. You only commit to one step at a time.
Which languages do you work in?
German, French, Italian, English and Spanish.
