Skip to content
Perilaw

PERI Labs · Document analysis

Large-scale document analysis with AI

From document collections to the questions of practice: organised sources, distinct evidence and critical points brought to light for assessment.

Discuss your enquiry ↗Start with the context and your objective, without confidential documents.

Document analysis method

Experience gained in due diligence and in managing complex case files has led to the development of tools for reading, connecting and verifying large document collections.

The same method applies to related needs: complex due diligence and civil or criminal fact-finding where facts, chronologies and references between documents have to be reconstructed.

Each adaptation is modelled on the documents, the legal questions and the checks the case requires. Source traceability and data protection are an integral part of the method, in the service of the professional’s assessment.

  1. The collectionDocuments from different sources and in different formats: an archive to make navigable, rather than a set of ready-made answers.
  2. The readingIdentify useful references, check uncertain scans and distinguish information that is actually documented.
  3. The connectionsReconstruct facts, parties, chronologies and relationships between documents, retaining the ability to return to the source.
  4. The reviewGaps, ambiguities and divergences are made explicit before a reading is used in legal work.
A loupe on a dense page with a few lines highlighted

Reviewing document analysis

Every step leading to the assessment remains traceable. The functions described are distributed across the various projects according to the scope of each, and not all are present in every tool in the catalogue.

Readability before interpretation
OCR check masks and review of uncertain readings make it possible to separate document content from acquisition errors.
Information requiring protection
Pseudonymisation and masking of personal and business data are handled with dedicated procedures, and the DPO’s assessment is integrated with the lawyer’s.
Sources, relationships and gaps
Sheets, chronological reconstructions and reconciliations distinguish what the documents support, what is missing and what requires further verification.
A printed report with pencil notes in the margin

A result that can be reviewed

An explicit scope
Documents examined, questions and unavailable information define the meaning of the analysis. Gaps in the collection remain explicit even after AI processing.
Visible issues
Source references, inconsistencies and open points make the work verifiable and correctable by the professional. Every summary can be checked against the documents and references used.
Example entry · fictional data

Fictional example showing the structure, not a real case or an output from a particular application.

Question
Term of an audiovisual licence.
Example documents
Contract Alpha, clause 4: expiry 31 December 2027. Schedule Beta, p. 2: expiry 30 June 2028.
Issue to check
The dates differ. Check which document prevails and whether there is a subsequent agreement; do not automatically infer an extension.

Comparing and reviewing results

In some projects, AI models from different vendors are compared to bring out disagreements and resolve them in structured review cycles; unresolved points remain visible and are submitted to professional assessment.

Distinguish different kinds of review

A second reading within the same system is not a cross-vendor review. Comparing models from different providers can reveal disagreements, but their agreement does not establish correctness.

In legal research, debating the same sources is not equivalent to independent research. Coverage, contrary authorities, relevance and links to the texts still require examination.

PERI Labs

Document-analysis projects

The document-analysis projects, from tools already in use to research prototypes: each entry describes functions and safeguards, with no promises of performance.

In testing

Legal Due Diligence

From case file to a structured analysis sheet following the firm’s working template, with every item linked to its source, gaps made visible and the questions to put to the professional.

Documented capabilities and limitations

In testing

Credit Position Analysis

Reconstruction of claims, guarantees and the history of a position from heterogeneous documents, with a single list of points to review, whether reading or privacy issues, and a second consistency check before the final sheet.

Documented capabilities and limitations

Prototype

AIgorà

Classification and comparison of large text collections against a common scheme, with assessments compared, disagreements made explicit and final decisions left to a person.

Documented capabilities and limitations

In testing

ARGO Admin – Case-law research

Guided research from an abstract question: complementary searches, supporting and opposing lines of authority, tracing of cited precedents, and an opinion accompanied by the sources and extracts to verify.

Documented capabilities and limitations