Zetamind · AI agents and company memory

What Zetamind does

Three building blocks, one system.

Zetamind is the system Cédric Atticot dit Ravino built and uses every day on his own servers. Each block is labelled: what already runs, and what is still being built.

zm-brain

Company memory

Your documents and conversations in one place. The goal: your teams find an answer in one search, and your AI agents use it within the limits you set.

Running for Cédric

Notes, documents and the history of conversations with AI tools, searchable with one question: “what did we do on this file?”
Encrypted backup every night, restore tested
Agents read and write in the same memory

In progress

Visual graph interface
Multi-user access with a personal login

Snapshot of 27 Sep 2026: 1,257 notes, 43,904 searchable messages.

zm-spirit

AI agents

Assistants that take on one precise task: sorting requests or drafting a reply. Each one only sees what it is allowed to see.

Running for Cédric

An assistant on Telegram, with her own skills
An AI router limited to the chosen models
A sorting agent evaluated on 19 cases: 14 correct decisions (26 Sep 2026)

In progress

A workspace for each user
Admin console
Connection to another team's memory

Your tools, connected

Automation

Connect your tools so data moves between them without copy and paste. Only then hand repetitive tasks to agents, one at a time, and measure the time saved.

Running for Cédric

Automatic nightly collection of the main server's history
One shared work thread across his AI tools, on his machines and on Telegram

Built by Cédric as a consultant

Consulting assignment for a regional health observatory: connection to open data and an engine that automates data processing and checks. 23 of 55 topics now update with no manual work; platform delivered in 2026.

In progress

Zetamind OS

The interface that will bring the memory and the agents together, with a login for each person. A first prototype has been running behind restricted access since 25 Sep 2026.

Method

Scorecard, deployment, training.

In that order. An agent plugged into messy information saves no one any time.

Know where to start

It starts with the online scorecard: 12 questions, 3 minutes, a score out of 12 and the area to work on first. If your project fits, a conversation follows to look at your tools, your repetitive tasks and your sensitive data.

Where your information lives, and who can access it
The repetitive tasks that cost the most hours
What must never leave the company

Take the scorecard

A memory, then a first agent

First, your tools are connected to a shared company memory with three access levels: public, internal, confidential. Then a first agent takes on one frequent, simple task, and the time saved is measured. The AI model, sovereign or not, is chosen to fit your constraints.

Your tools stay in place: CRM, email, calendar, invoicing
One simple indicator: hours saved per week

Your teams, on their own tasks

Since November 2025, Cédric has trained developers from companies, SMEs and non-profits in generative AI. Training starts from your real cases.

Generative AI for software development (GitHub Copilot, ChatGPT)
The augmented developer, with Claude Code
A one-page AI usage rule, written with you
A measurable 90-day goal

Data flow

From your tools to your teams.

An agent is only as good as the information it receives. Every deployment follows this path, and each step has its rules.

01 · Input
Your tools
Email, documents, CRM, calendar, invoicing. Nothing is replaced.
02 · Memory
Company memory
Everything is indexed and backed up. Each piece of information gets an access level: public, internal or confidential.
03 · Action
AI agents
They search, sort, draft and prepare. They only see what is allowed.
04 · Decision
Your teams
They review and decide.

1,257 notes in Cédric's memorySnapshot of 27 Sep 2026
43,904 messages from conversations with AI tools, findable in one searchSnapshot of 27 Sep 2026
Encrypted backup every night, restore testedTest of 26 Sep 2026

AI scorecard

Is your company ready for AI agents?

12 questions, 3 minutes. You get:

Your AI readiness score out of 12
The area to work on first: memory, processes or teams
3 concrete recommendations for your situation

Free · Contact details requested at the end to show your results

Take the scorecard

Question 1 of 12 · Company memory

Your key documents (procedures, contracts, offers) are stored in one place that the whole team uses.

Yes No

Answering here starts the scorecard.

Who is behind it

Cédric Atticot dit Ravino

AI solutions engineer. 15 years in digital and AI, always between business teams and tech. He designed Zetamind and uses it every day.

cedricatticot.com LinkedIn

2011 to 2018
VINCI Immobilier. Digital projects lead. More than 300 platforms, a data warehouse and dashboards for the executive committee. 129 new wealth-management adviser partners from a single campaign.
2021 to 2022
Longchamp. Web development lead. Complete redesign of the e-commerce site and its Salesforce platform, with new payment methods.
2022 to 2024
Louis Vuitton, via Capgemini. Apple Pay and Google Pay across Cybersource, Adyen and Stripe, with IT, legal and marketing teams in several countries.
2024 to 2025
L'Oréal, via Smartch. Scoping, specifications and vendor follow-up; workshops and user training.
2025 to 2026
A regional health observatory. Data platform delivered in 2026: 23 of 55 topics now update with no manual work.
Since 2025
AI trainer. For companies, SMEs and non-profits.
2015 to 2017
ESSEC. Master's in Information Systems Management.

Contact

Let's talk about your company.

The simplest way: take the scorecard. You can also call or email Cédric directly. Paris and remote.

Take the scorecard