Traceable differences between documents

How can you obtain a reliable, sourced comparison without confusing contradictions, omissions and differences in scope?

1250 € per session, excl. VAT

In short

In a half-day session of 3 hours learn to use generative AI to compare multiple documents, identify where they align and characterise their differences. You produce a comparison matrix and a verifiable synthesis, then measure their quality by checking references, omissions and interpretations.

Why this training?

A fluent synthesis can conceal an exception, equate lead times expressed differently or attribute information to the wrong document. Comparison requires common criteria and precise references, not just a summary of each text. AI helps organise information and suggest connections. The participant remains responsible for checking cited passages, flagging missing information and not resolving a contradiction without documentary evidence. Ongoing case study: The fictitious organisation Avelune Équipements is preparing a meeting on its delivery terms. Three documents are available: a sales information sheet, terms of service and an update note. Lead times, exceptions and responsibilities differ; established differences must be distinguished from points requiring confirmation. Transfer to the workplace: In the following week, reuse the framework on a small, authorised set of documents from your organisation. Check each significant difference against the sources and record the necessary corrections before sharing the synthesis.

Objectives of this training

  • Define the document set, its versions and the comparison criteria
  • Formulate a request for structured extraction with source references
  • Classify differences as contradictions, omissions or variations in scope
  • Write a synthesis distinguishing verified facts from points requiring confirmation
  • Measure coverage of the criteria and accuracy of the references

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Assistants and coordinators preparing syntheses from multiple documents
  • Research officers and information specialists carrying out routine document comparisons
  • Team leaders who need to explain differences between versions or formats

Prior skills

  • Have prior experience of making simple requests in a generative AI assistant
  • Be able to locate a version, a date and a section in a document

What you will work on

  1. Define a basis for comparison

    Objective: Make the three documents comparable without erasing their differences in scope. Inventory, versions, effective dates, intended audiences and units; selecting criteria and identifying incomplete texts.

  2. Extract with evidence

    Objective: Obtain structured information that can be located in the sources. Instruction specifying the document set, criteria, format and a prohibition on filling in missing information; short quotations and section references; processing excerpts where necessary.

  3. Characterise the differences

    Objective: Separate genuine differences from misleading connections. Contradiction, omission, rewording and difference in scope; attention to working days, exceptions and chronology.

  4. Check and report

    Objective: Deliver a synthesis whose reliability is explicitly assessed. Coverage of expected cells, accuracy of references and comparison against the training framework; separation of findings and interpretations.

What you take away

  • Scope sheet and source inventory
  • Reusable instruction for comparative extraction
  • Matrix of similarities and differences with references
  • Decision-making synthesis and quality control record

Method and assessment

  • Short teaching inputs followed by immediate practice on the ongoing case study.
  • Individual work followed by work in pairs, with peer review of the outputs.
  • Guided group discussion: what works, what needs correcting and what remains a human decision.
  • Preparation: Ensure access to an AI tool authorised by the organisation; check the permitted methods for transmitting texts. Prepare the fictitious training document set and its reference framework; any material contributed personally must be anonymised and authorised.

Observable criteria: Each source has an identifier, a version or an explicit statement that the version is unknown. ; Each expected cell contains sourced information or the statement « not found ». ; Each significant difference is characterised and checked against the relevant passages. ; The synthesis never turns missing information into a contradiction. ; The check quantifies cell coverage and the proportion of accurate references, before and after correction..

Frequently asked questions

Do you need to submit the complete documents?

Only if authorised. Identified excerpts may suffice, provided that the limitations of the document set are stated.

Can scanned files be compared?

Yes, if their text is usable; first check the transcription, particularly tables, figures and notes.

Does AI decide which document is authoritative?

No. Without an explicit priority rule, it must flag the conflict and request human validation.

Are only successive versions compared?

No. Documents of different types are suitable if their common criteria and scopes are clearly defined.

Last updated: