Corpus and traceable synthesis

How can you turn scattered documents into a verifiable synthesis without confusing sources with AI interpretations?

1250 € per session, excl. VAT

In short

In a half-day session of 3 hours participants define a document review question, select a corpus and build a reading matrix using an authorised generative AI tool. They produce a synthesis with source references, distinguish areas of agreement from contradictions, then measure its traceability and coverage of the defined scope.

Why this training?

An accumulation of summaries does not constitute a document review: without an explicit scope or selection criteria, documents that are easy to find take precedence over relevant sources. Contradictions then disappear behind a synthesis that suggests an artificial consensus. AI can help suggest classification categories and extract comparable elements, but its wording must be checked against the documents. The method maintains a verifiable link between each conclusion, its source passage and the limitations of the corpus. Continuing case study: The fictitious organisation « Avelune Médiation » is considering extending the opening hours of its reading spaces. A training pack of eight fictitious documents brings together user feedback, internal notes, studies and conflicting arguments. Management expects a synthesis that distinguishes documented benefits, constraints and questions that remain open. Transfer to the workplace: Within the following two weeks, reuse the matrix for a document review question within your organisation, using authorised data. Compare theme coverage and the proportion of traceable claims before and after correction, then retain the useful AI instructions.

Objectives of this training

  • Formulate a document review question and its temporal and thematic boundaries
  • Select documents using explicit relevance and reliability criteria
  • Build an extraction matrix linked to source passages
  • Write a synthesis that distinguishes findings, divergences and gaps
  • Measure the traceability of claims and correct insufficiently supported conclusions

Who is it for and what are the prerequisites?

Roles and tasks concerned

  • Research officers and information specialists preparing synthesis dossiers
  • Project managers who need to inform a decision using heterogeneous sources

Prior skills

  • Have previously formulated and refined a simple request to a generative AI tool
  • Be able to identify the author, date, title and references in a document

What you will work on

  1. Define the scope of the question

    Objective: Turn a broad request into a workable scope for a document review. Central question, sub-questions, period, document types and exclusion criteria; distinction between a targeted review and a systematic review.

  2. Assess and organise the corpus

    Objective: Make document selection choices explicit. Provenance, currency, method, author interests, duplicates and identifiers; search log for future additions.

  3. Extract without losing the evidence

    Objective: Compare sources using a common framework. Reading matrix columns: theme, finding, method, limitations, quotation and location. AI instruction: closed corpus, missing information flagged, no references added.

  4. Synthesise and audit

    Objective: Produce a usable report with visible limitations. Thematic structure, areas of agreement, contradictions and uncertainties; calculation of traceability and coverage rates.

What you take away

  • Scope sheet and selection protocol
  • Corpus register with inclusion and exclusion decisions
  • Reading matrix with references and supporting passages
  • Synthesis report accompanied by its completed checklist

Method and assessment

  • Short instructional inputs followed by immediate practice on the continuing case study.
  • Individual work followed by work in pairs, with peer review of outputs.
  • Guided group discussion: what works, what needs correcting and what remains a human decision.
  • Preparation: Arrange access to a generative AI tool authorised by the organisation; only fictitious or anonymised documents will be used. Read the case study question and browse the training pack; identify a document whose relevance seems questionable.

Observable criteria: The question specifies the audience concerned, the themes and the period covered. ; Each document included or excluded has a recorded justification. ; Each factual claim in the synthesis points to an identifiable passage; the traceability rate is calculated. ; Divergences are attributed to the relevant sources, without inventing a judgement to resolve them. ; Each planned theme is documented or flagged as a gap; coverage is calculated..

Frequently asked questions

Is an Internet search required during the session?

No. The supplied corpus enables participants to practise selection and synthesis; further research requirements are recorded.

Can the AI add references?

Any suggested reference outside the corpus remains a lead to be verified and is not automatically included in the synthesis.

Can internal documents be used?

The exercises remain fictitious or anonymised. Subsequent application requires compliance with confidentiality rules and the use of authorised tools.

Will the review be exhaustive?

No. The deliverable is a targeted review with an explicit scope, source selection choices and gaps.

Last updated: