We lead companies into the age of artificial intelligence.

How we work

An engineer assigned to your company
The same person carries diagnosis, going live, training and keeping it running.
What we install belongs to you
The software, the developments that integrate it into your systems and the documentation are not proprietary: they can be taken over by your own teams, or by another provider's.
A system reliable and secure by design
A person can be included in the automated process to validate the sensitive steps. Answers come from verifiable sources. Spending is capped at the amount you set, and the system's access to your tools is defined one tool at a time.

Software

We examine software as it appears, we put it through its paces on our own machines, and we keep only what proves reliable in real service.

We prefer open source software whenever it meets the need. We look into software outside the catalogue case by case.

We develop the layer between that software and your systems: the connection to your tools, the rules of your trade, the migration and digitisation of your data.

We install no software whose business model is your data.

The full catalogue

Use cases

We offer four forms of engagement, and they can be combined. The diagnosis establishes which apply, and in what order.

The augmented workstation
We give your teams tools they use every day: a conversational assistant, a search that answers from your own documents and shows its source, and assistance directly inside their office software. These tools are wired into your systems: they sort incoming mail, and create the tasks in your teams' tracker.
The missing leverage
Many companies already pay for AI assistant licences, but their teams only use them to draft. We take those subscriptions and add what they lack: reusable procedures, connectors to your internal tools, and the links that carry information from one piece of software to the next.
Automating a use case
An automation can run entirely on its own, or be broken up by human steps wherever that is called for. We begin by establishing what the task actually does, we redesign the process when it does not deserve to be automated as it stands, then we put it into service.
Support with digitisation
A process can only be automated if the documents it handles are readable by a machine. We first take in hand whatever is not: paper files, archives scanned without text, scattered spreadsheets.
Use cases in detail

Control

An AI system can make mistakes. Avoiding them means identifying the sensitive steps at design time, and placing the controls there.

A human in the loop
We define together the moments where a person validates. The file comes back to them with what the decision requires, and the steps concerned are written into the scope before the work begins.
Answers you can check
A model queried on its own invents. The system answers from your documents, gives its source, and flags the questions it could not answer.
Capped costs
Every installation includes a gateway, and no model call bypasses it. You see spending use case by use case, and you set the ceiling.

Hosting

You decide where your data is processed, and we install the system there.

A server on your premises
Your documents are processed on your own network, and no third party can receive an access request. The server can be leased, or bought by you before the work begins.
Our hardware, in Lausanne
Your data stays in Switzerland, under Swiss law, in an installation shared with no other client.
A Swiss host
Swiss law, and no company subject to United States jurisdiction.
A European host
European law, and the GDPR.
An American provider
The CLOUD Act lets a United States authority request access to your data from the provider, wherever the servers are.

If you have no hosting, we can provide it on our own hardware, in Switzerland. If your data must not leave your premises, we can install a server there.

Hosting in detail

Offer and price

Each stage is decided separately, at a price settled before it begins. None is billed by time spent. They hold for all four forms of engagement, even if a given engagement does not call for all of them.

What it producesWhat we commit to
01Diagnosis
We put a figure on what your repetitive processes cost today, and we name those where automation would pay off most.
It is not billed, and you keep the document, whether you go further or not.
02Going live
We put one use case into service, on a date agreed with you.
The price is settled in advance and falls due only on going live. If the date is missed, we do not bill it.
03Training
Your teams learn to run the system, and to handle the files it sends back to them.
It is included in the price of every installation, and we also give it on its own.
04Keeping it running
We supervise the system, we keep the software up to date, and we widen the scope as your work changes.
A monthly subscription, cancellable at any time.

The scope and the definition of “in production” are written and signed before the work begins.

Team

A team trained at EPFL: two artificial-intelligence engineers, one of them a doctor specialised in private systems, and a projects lead who accompanies your teams from defining the need to taking the tools in hand.

Albert Troussard
Artificial-intelligence engineer
Elena Dimitratchkova
Projects and client relations
Maxime Perrot
Artificial-intelligence engineer
The team

Get in touch

Thirty minutes, by video call.

We look at your documents and say what we know how to do, including when the answer is no.

Pick a date

+41 78 305 23 12
bonjour@advazia.ch
Avenue Édouard-Dapples 21
1006 Lausanne, Switzerland

In writing

A person reads your message, and replies within 24 hours.

✓ Message received. We reply within 24 hours.
Book a call