The Marylink vocabulary, explained simply.

The AI Act classifies AI uses by risk level and mandates transparency and oversight. Marylink's governance (roles, validations, traceability) helps you document and steer your compliance requirements.

An AI agent carries out tasks autonomously. Connected to Marylink, it relies on validated practices rather than improvising, which makes its actions governed and traceable.

The AI diagnostic places your organization on the depth and speed of its usage, then proposes prioritized actions. It's the starting point before any Marylink rollout.

Augmented creation means producing with an AI that knows your context: your practices, spaces and standards. The output reflects your context rather than a generic one.

B2B2B lets a firm turn its know-how into an offering: capitalize a method in Marylink, then make it available to clients under your own brand. The practice becomes a product.

The canvas is Marylink's creation surface: you compose content from existing blocks and practices, letting AI execute your team's know-how rather than a generic prompt.

Without capitalization, AI makes you execute faster while eroding institutional memory. Capitalizing means keeping the executable trace of what worked, so the organization grows richer instead of emptier.

The champion is a space's human engine: encouraging contributions, highlighting good practices and keeping momentum. Without a champion, a space falls asleep.

CREW (Concepts, Roles, Environments, Workflows) is the governance grammar that structures Marylink: who is responsible for which practice, where, and through which steps. Published by M. Elmoukhliss and H. Mary.

The dashboard turns activity into steering. You see what is used, what is slipping and what needs action, to monitor and guide AI usage.

GDPR governs the processing of personal data in Europe. Marylink builds in its requirements by design: European hosting, minimization, traceability and data-subject rights.

In Marylink, governance isn't a brake: it's what makes a practice trustworthy. Space roles, validation steps, expert reviews and revisions ensure you reuse what's sound, not what's «it depends».

Knowledge collapse, formalized by Acemoglu, Kong & Ozdaglar (NBER, MIT), names the impoverishment of collective knowledge in the AI era. The authors point to aggregating and governing know-how as a way to preserve performance and resilience.

MCP lets an external AI assistant securely read and act inside Marylink: find a practice, publish, review. Your practices become tools the AI can use, no copy-paste.

The moderator keeps a space healthy: arbitrating, correcting, archiving what no longer belongs and enforcing the space's rules.

Marylink is modular: each capability is a module you turn on or off without touching the rest. You only pay for, and show, what you use.

Packs simplify the choice: instead of ticking dozens of options, you enable a coherent scope. Start from a Core base and add packs as needs grow.

A practice is the atom of the Practice Graph. Rather than dead text, it's an assembly of typed blocks (content, prompt, idea, style) you can reuse, evolve and have an AI execute.

Practice Engineering is the scientific approach behind Marylink: formalizing what people actually do well, so it becomes a practice anyone (and any AI) can run. It's the shift from knowledge you document to practice you execute.

The Practice Graph is the heart of Marylink: it turns your organization's know-how into a network of typed, connected and reusable practices. Unlike a static wiki, it grows with every use and can be executed by humans and AI agents alike.

A publication is the shared form of a practice. It carries a version history, expert reviews, revisions and a validation path, so quality is steered, not left to chance.

RAG anchors AI answers in your documents and practices. The result: fewer hallucinations, sourced answers, and an AI that speaks about your organization rather than the web at large.

Reuse is the return on know-how: a practice captured once, executed dozens of times. Marylink makes it visible and measurable («reused ×7»).

A review turns intuition into a trust signal. An expert scores the practice against explicit criteria, so others (humans and AI alike) can reuse it knowingly.

A seat is one active user. Add or remove seats as your teams change: billing follows real usage, with no rigid commitment.

At Marylink, sovereignty is concrete: data hosted in Europe, never used to train third-party models, within a GDPR- and AI Act-compliant framework. Your know-how stays yours.

Spaces organize know-how by domain. Each has its members, roles, pins and recommendations: what works rises to the collective, the rest stays out of the way.

Space roles distribute responsibility for practices. They define who can publish, who reviews, who moderates and who animates: the «Roles» brick of the CREW framework.

Validation steps make a practice's workflow concrete. Each column has its rules: until they're met, you know exactly why a practice is blocked: the «Workflows» brick of CREW.