Assistant for internal documents

How to build an AI assistant based on your internal documents?

  • ChatGPT
  • Claude
  • Microsoft Copilot

1960 € per session, excl. VAT

In short

This program is for support, quality, knowledge management teams, and AI leads. Participants organize a document corpus and configure an assistant, using tools like ChatGPT, Claude, or Microsoft Copilot. They leave with a prototype and reliability tests, enhancing the precision of responses grounded in their official sources.

Why this training?

The problem
Teams struggle to effectively leverage the vast amount of information contained within their internal documents.
What it costs today
Employees waste time searching for information, answers to questions are inconsistent, and decisions lack precision.
Why this skill
This skill is necessary to build reliable tools that draw upon the company's official sources.
What you gain
Participants and the company gain autonomy in configuring assistants based on ChatGPT, Claude or Microsoft Copilot, quality in the answers obtained, and reliability in the information disseminated.

Objectives of this training

  • Precisely define users, key questions, and authoritative documents for an assistant.
  • Clean up versions, titles, metadata, and access of source documents.
  • Configure instructions, citations, and handling of missing information for an assistant.
  • Measure the relevance of sources and the fidelity of an assistant's responses.
  • Define the roles and processes for updating and human validation of an assistant.

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Customer support teams
  • Document quality managers
  • Knowledge management specialists
  • Artificial Intelligence referents
  • Information managers

Prior skills

  • AI Fundamentals and AI Prompting completed, or equivalent experience (a learning prerequisite, not a registration condition).
  • Understanding of challenges related to internal company data
  • Knowledge of the types of official documents within the organization
  • Experience in organizing and searching for documentary information
  • Familiarity with content or knowledge management
  • A paid subscription to an AI solution offering the features used, the organisation’s authorisation for third-party tools, and the actual accounts and access rights.

To prepare before the session

  • A computer per participant
  • Access to selected tools such as ChatGPT, Claude, or Microsoft Copilot
  • A business case to work on with fictitious or anonymized data

What you will work on

  1. Choosing the Right Scope

    This module allows defining precise user needs and relevant documentary sources.

  2. Preparing the Sources

    This module allows organizing and structuring the document corpus for optimal utilization.

  3. Building the Prototype

    This module allows configuring an assistant and testing it on various queries.

  4. Evaluating and Maintaining

    This module allows ensuring the quality of the assistant's responses and planning its evolution.

What you take away

  • An organized pilot corpus
  • A prototype document assistant
  • A set of reliability tests
  • An elaborated action plan
  • Practical support materials and completed templates

Method and assessment

  • Demonstrations with explanations using a concrete case
  • Guided exercises followed by application to your role
  • Peer review of results and source verification
  • An individual or team action plan

Assessment is formative: participants complete and review a practical case linked to the intended deliverable (An organised pilot corpus, a document-assistant prototype and reliability tests). Criteria include meeting the brief, accuracy, checks performed and the ability to explain decisions. Each participant identifies an application to test in their own work.

Frequently asked questions

What concrete outcome can my team expect after this session?

Your team will be able to organize a relevant document corpus and configure an assistant prototype based on your internal documents. They will leave with a set of tests to evaluate the reliability of the answers, which will improve the accuracy of disseminated information.

Who should participate in this session and what are the prerequisites for getting the most out of it?

This session is intended for support teams, quality, knowledge management, and AI referents. Participants must understand the challenges related to internal data, know the official documents, and have experience in document organization. No specific technical level is required for the tools.

What types of data or documents will we use, and how is confidentiality ensured?

We will use a business case to work on, ideally with fictitious or anonymized data. The use of real non-sensitive data is possible if it has been prepared in advance. The session emphasizes methods for protecting and properly managing information.

What kind of practical exercise will be carried out during the session?

Participants will organize a document corpus and configure an assistant prototype using tools like ChatGPT, Claude, or Microsoft Copilot. Reliability tests will then be conducted to evaluate the accuracy of the answers obtained.

What are the limitations of the tool or approach, and who validates the final result?

Assistants based on internal documents have their limitations, particularly regarding missing information or topics not covered by the corpus. The final validation of the results and answers provided by the assistant rests with the business experts and the organization's management.

How does the session content adapt to my organization's business and specific cases?

The session is designed to be applied to a specific business case chosen by the organization. Discussions, exercises, and the action plan are geared towards the documents and issues specific to your context. The goal is to prototype an assistant that meets your needs.

Last updated: