Agentic orchestration · AI Act compliant
Edugent is the layer that orchestrates AI agents working for your institution and keeps you in control of quality, costs, security and compliance. Agents handle the volume; decisions always remain human.
Agentic orchestration
Between your teams, your systems and the agents, Edugent sets the rules and enforces them, in every module.
01 · Quality
When everyone uses AI in their own way, with no set process, results vary from one person to the next and from one day to the next. With Edugent, every task follows a defined workflow: agents draw on your institution's knowledge base, each staff member has a digital twin calibrated to their voice that learns from every correction, and sensitive topics go to a person, with no generated reply.
02 · Costs
Where AI costs are usually unpredictable, a single unit: the credit. 750 per week per module (30,000 per year), in a shared reserve. Limits per institution and per user, checked before every run; your administrator is alerted before they are reached. Never a surprise invoice.
03 · Security
Each institution is kept separate. A conversation can only be read by its participants: your administrators see usage (who, which module, how many credits), never the content. They can pause or stop any agent at any time.
04 · Compliance
Decisions always remain human (Art. 14). When a twin deals with a third party, this is flagged by default, with direct access to the person (Art. 50). Audit log retained for 1 year. Data stored in the European Union, optional 100% EU AI processing.
Edugent is compatible with Claude, GPT, Mistral and its own instances hosted by OVH in France.
The product in pictures
Your first module live in four weeks. The twin drafts in the staff member's voice, from your sources. Decisions always remain human: approval from a phone, transparency for students.
Connected to your tools, with no migration: Microsoft 365 (Outlook, SharePoint, OneDrive), Google Workspace, any IMAP mailbox, CalDAV calendars, Salesforce, SAML 2.0 SSO. The rest of your IT systems connect through our API, scoped with you at onboarding.
See the platform →To approve · 12 drafts
Sophie Martin · assistant 78% calibrated
Yuxin Zhao
Re: Deadline
Hi Yuxin, I'm aiming…
Carlos Mendes
Re: Application
Hello Carlos, OK…
Aisha Bakr
Re: Eiffel Scholarship
Hello Aisha…
Wei Liu
Re: Rejection
Sensitive topic · blocked…
Application pile · MSc Finance 2026
127 applications · pre-scored on your committee criteria · ready for review
Yuxin Z.
MSc Finance · China
Aisha B.
MS Marketing · Morocco
Carlos M.
Global MBA · Brazil
Rohit K.
MSc Finance · India
The score is a reading aid. The committee remains the sole judge: Article 14 of the AI Act, by design.
In every module
Reading applications, pre-scoring, a one-page brief for the admissions committee: the agent prepares, the decision stays strictly human. The same approach applies to student records, internships, mobility and accreditations.
One foundation, three families
01
Three families of modules, shown below: Fill, Run, Raise.
02
Each module is set up around your processes: steps, instructions, documents, approval levels.
03
We design it with your teams, on the same foundation: the same approval rules, the same credit reserve, the same audit log.
Family 01
Student numbers
For: Heads of Admissions / Recruitment
Family 02
Efficiency and compliance
For: Registrar / CFO
Family 03
Reputation and funding
For: President's office / Research Director
Each module's impact reference points are indicative and calibrated on your figures.
Six months to measure, on your own figures, what orchestration changes for your teams, under their constant oversight.