Skip to content
ETHERIA Conseil

Generative AI in production

RAG agents, MCP, n8n and governance of AI services (Microsoft Foundry, Gemini).

Format: a 10-day assessment, then industrialisation

The problem

Your generative AI proofs of concept are not reaching production: security or data questions are blocking the projects. Or AI workloads are already running, with no tracking of their costs or of the quality of their answers.

What I do

  • I assess your platform from an AI perspective: identity and access to models, private networking, data governance, integration with your systems through APIs and MCP.
  • I design RAG agents and connect them to your tools through MCP or n8n.
  • I set up the governance of AI services: access controlled by policies, models deployed on Microsoft Foundry or Gemini Enterprise Agent Platform.
  • I make the agents operable: identity, private networking, secrets management, tracking of costs and answer quality, operational runbooks.

Possible formats and deliverables

  • AI readiness assessment10 working days

    • How ready your platform is.
    • 2 or 3 realistic use cases, prioritised.
    • A target architecture.
    • A monitoring setup: cost and drift indicators, alert thresholds, recommended tools.
    • The path from proof of concept to production.
  • AI industrialisation

    • A secure service (identity, private networking, secrets management), observable and with costs under control.
    • Its integration with your systems.
    • Its operational runbooks.

My stance

AI starts with the platform: identity, networking, data governance and cost control decide whether an AI project succeeds, long before the choice of model.

Track record

  • Three generative AI agents in production in a large international retail group:
    • a RAG agent that searches more than 6,000 IT contracts and answers in about a minute instead of several hours;
    • an n8n agent that helps incident managers outside business hours;
    • an agent for the platform’s users, with support requests down 30% in two years.
  • Access to AI services controlled by policies; models on Microsoft Foundry and Gemini Enterprise Agent Platform; n8n agents in Docker on GKE.
  • AI Champion of the Infrastructure department: weekly sessions with the development teams on coding assistants (Gemini Code Assist).

Contact

A context close to yours?

Describe your context and I will get back to you within 48 business hours.

Describe your context
Location
Paris