Technologies
We sell no licenses and are not a reseller for any vendor. The choice of tool depends on the use case, the required level of confidentiality and the budget. Here are the ones we use in our engagements and training.
Our foundation
Each tool has an area where it outperforms the others. We train your teams to know which one to open.
Our go-to tool for long documents and professional writing in French. Large context window, controllable tone, reliable adherence to formatting instructions.
The most widely used tool across teams, so training often starts here. A strong all-rounder with an ecosystem of extensions and internal GPTs.
Built into Google Workspace, which matters for organizations whose documents and email already live there. Strong image and video processing.
An agent that works inside a code repository and carries out tasks end to end: reading existing code, modifying it, testing. It is what we use to build most of our automations.
A second development agent, used in parallel on long projects and to compare two approaches to the same problem before deciding.
A model with published weights, so it can be installed on your own servers. The right choice when data cannot leave the company or when per-request cost must be kept under control at high volumes.
The names and logos mentioned belong to their respective vendors. Their presence indicates use in our engagements, not a commercial partnership or certification.
Method
Data sovereignty
For regulated matters — banking data, health data, contracts — we deploy an open-weight model on your infrastructure or with a hosting provider you control. No data leaves the perimeter.
The trade-off is made explicit from the scoping stage: reasoning quality below the latest proprietary models, in exchange for full control over how documents flow.
A one-hour assessment is enough to decide between an off-the-shelf assistant, an agent connected to your systems and a model hosted on your premises.
