root@fr:~$ connecteurs/outlook
Your Outlook inbox, sorted and prepared. Before your first coffee.
An agent that reads what its scope allows, sorts, prepares sourced drafts and the file for your next meeting. It sends nothing: you sign. On your instance, in France.
What the agent prepares
The inbox work that does not need your judgement, already done.
The morning sort
Threads classified by what they expect from you: to decide, to answer, to read. The rest is filed, not deleted.
Draft replies
Prepared with the thread's history and attachments, in your tone. They wait as drafts: nothing leaves without you.
The meeting file
Before every meeting on the calendar: the latest exchanges, the commitments made, the documents cited. Five minutes of reading, not thirty of searching.
Follow-ups
Threads left unanswered beyond your deadline, listed with their context and a follow-up draft ready to sign.
Long-thread summaries
Forty messages summarized with links back to the originals: who said what, what remains open, what falls to you.
The rules
An agent in the mailbox, under IT's rules.
The agent connects to Outlook through the Microsoft 365 API, with the rights your directory grants it. Nothing to install on workstations, nothing to change for your teams.
Scope per mailbox
The agent reads the mailboxes and folders its scope declares, nothing else. Access goes through your directory and is revoked like any other right.
Audit log
Every read, every draft, every denial: a timestamped line, queryable, exportable to your SIEM.
Never an autonomous send
The agent produces drafts. Sending is yours: an email binds the company, and that signature cannot be delegated.
Your email stays in France
Processing happens on your instance, not on a consumer AI platform. Nothing trains a public model, nothing leaves your jurisdiction.
Frequently asked questions
What we get asked about Outlook.
Can the agent send an email on its own?
No, by design: it prepares drafts and lists, and sending requires human validation. That is an architecture rule, not a setting someone can forget.
What exactly does it read?
What its scope declares: these mailboxes, these folders, this much history. The scope is decided with your IT department, shows in the audit log, and is revoked through your directory.
Do we have to change mail systems or install an add-in?
No. The agent connects through the Microsoft 365 API, server side. Your teams keep Outlook as it is, on all their devices.
How is this different from the AI embedded in US office suites?
Ownership and jurisdiction. Here, the AI layer runs on your instance in France, under your audit log, and your email never trains a third-party model. The assistant embedded in a US suite is a service you rent, under its law, not yours.
The demo runs on a test mailbox. Not on slides.
In one workshop, we plug the agent into a demo mailbox, and you watch the sorting, the cited drafts and the log. Only then, your real scope.