Create Lucy, an Azure OpenAI chatbot on 1.5 million GED documents

What was done

Le Journal du Net (Xavier Biseul, September 17, 2026) questions Jean-Daniel Delaplagne, head of IT Project Tools Section of ITER (merger reactor, Saint-Paul-lès-Durance, ~2,000 employees). Lucy connects on Azure Specific agents have also been developed for small corpuses (100 to a few thousands documents). 400 Microsoft 365 A monthly committee reviews IA projects. ITER has the status of an international organization and Microsoft Frontier Firm. This box isolates Lucy/GED, not the 300

Before you start

Tool cited by the source: Lucy (Azure OpenAI); other tools cited: Azure OpenAI and Microsoft 365 Copilot. Fill in the bracketed items: [coller], [liste ou extrait]. Strip confidential data that is not needed and check your tool’s access terms. Features and plans can change.

The method, in three steps.

1. 01

   Gather the context

   Spot the bracketed items in the prompt ([coller], [liste ou extrait]) and prepare what they ask for.

2. 02

   Adapt the prompt

   Copy the prompt below, replace the bracketed items, then paste it into Microsoft 365 Copilot.

3. 03

   Reread before using

   Reread the result, correct it if needed and keep the final decision. Caveats are detailed below.

Your starting point.

Replace the bracketed items, then adapt the proposal to your situation. This prompt is written by the iadujour desk It is not a quotation from the source.

Technical question (without export-control secret): [coller]. Authorized documents: [liste ou extrait].

Answer with quote (title, date, extract). If nothing in the corpus, say it. Do not mix two revisions of the same document.

What deserves your attention

Useful skills

Official skills on the same task, paired by the desk: they are not part of the described case.

• Obtain instructions to remedy a recommendation AdvisorMicrosoft · This skill helps to obtain rectification instructions for an already selected Azure Advisor recommendation, so that the client can execute them himself.
• Build or diagnose an Azure AI Search, Speech, OpenAI or Document Intelligence applicationMicrosoft · This skill guides the construction and troubleshooting of Azure AI Search, Speech, OpenAI and Document Intelligence applications.
• Recommend an Azure skill adapted to the taskMicrosoft · This skill is used for searching the Azure skills catalogue, to recommend an installable agent skill that corresponds to an Azure task: list plugins and skills from the API GitHub repository `microsoft/azure-skills`, read descriptions, find the plugin name in `plugin.json` (which may differ from the folder name), and then propose installation instructions depending on the client (Copilot CLI, Claude Code, or other).
• Build, deploy and operate Microsoft Foundry agents and modelsMicrosoft · Covers the full life cycle of Microsoft Foundry agents, models and resources: construction, deployment, evaluation, optimization, refining and operation, including azd, hosted agents, tools, batch or continuous evaluations, IC/CD, agent optimizer, model deployment and refining (SFT, DPO, RFT), RBAC, quotas, regions and knowledge indexes.
• Check and manage quotas and Azure capacityMicrosoft · This skill is used for consultation and demand for Azure quotas, as well as for routine use, to plan a deployment or choose a region: check service limits, use, and request an increase — this demand is free: only the resources actually used are paid, not the quota allocation.