A team of AI agents. One person who decides.
At Domaine du Net, part of the work is done by artificial intelligence agents: they write code, draft messages, track search ranking and review contracts. They are not people, and none of them decides in Julien’s place. Here is who does what, and above all what each one is not allowed to do.
The three rules that never change
Nothing risky without a yes.
An agent proposes, Julien approves.
The one who writes is not the one who reviews.
A model that has just built a solution defends it. That is why review is handed to another agent.
AI prepares, a human approves.
AI prepares the follow-ups and the messages to send; a human approves before anything is sent.
The team
The team, and each one’s limit
Each has a precise job and a written limit. The limits matter as much as the skills.
Liz
Chief of staff
AI agent. Receives Julien's topics, frames them, hands the work to the other agents, checks the result and prepares decisions.
Its limit: Never decides in Julien's place and does not do the work handed to the other agents.
Supervised by Julien
Marc
Business development & partnerships
AI agent. Drafts follow-ups and outreach messages, and analyses how prospects move through the funnel.
Its limit: Sends no message: Julien reviews and sends.
Supervised by Julien
Hugo
Developer
AI agent. Writes the code of our products and submits changes for review.
Its limit: Ships nothing on its own: every change is reviewed and approved before release.
Supervised by Julien
Nora
Acquisition & SEO
AI agent. Tracks how visible our sites are in search engines and spots content to create or fix.
Its limit: Changes no site: its recommendations are validated before publication.
Supervised by Julien
Vincent
Finance director
AI agent. Follows invoicing and payments, matches invoices to payments and prepares a regular financial update for Julien.
Its limit: Does not keep the official accounts and executes no payment.
Supervised by Julien
Adèle
Legal & compliance
AI agent. Reviews our contracts and terms of sale, and flags what is missing or out of date.
Its limit: Drafts no final contract and gives no legal advice: every clause is validated by Julien.
Supervised by Julien
What it looks like in your tools
Support
An AI assistant answers your customers on your website and on WhatsApp, from your knowledge base. Requests are tracked as tickets.
Pulse
An AI gateway for tools: generation, extraction, summarisation, embeddings and crawling, through a single API.
Video
Short marketing videos assisted by AI: script, scene editing, MP4 rendering.
Your website
A CMS that AI agents can drive to build a structured editorial site, which you then manage without a developer.
The risks of AI, and what contains them
Every risk has its safeguard. Here are the main ones.
AI can produce wrong or approximate code
Automated tests, then a review by another agent, before anything goes online.
A change breaks an existing feature
Regression tests, run on every change.
Moving too fast and breaking production
Nothing is merged without a merge request, a review and a green pipeline.
Approving a feature “on trust”
A preview environment you can open before going live.
Depending on a single AI provider
A multi-model AI gateway (Pulse).
AI makes up a plausible but wrong answer
Answers drawn from your knowledge base, within a limited scope.
What we do not do
Nothing is sent until a human has approved it. We do not let an agent commit the company. Nothing goes online without review. And we do not present as a person what is an AI.
If you want an assistant like this in your own organisation, we start with one day.
Book a call