How Clauseflint Reviews

Documents are received via secure upload or direct DMS connection. Supported formats include Word (.docx), PDF (text-layer), and RTF. Password-protected PDFs require unprotected submission.

Each document receives a unique ingestion identifier that persists through the review process and into the audit trail export. Batch submissions are accepted for deal packets; up to 200 documents per pilot engagement.

The document structure is parsed to identify section hierarchy, exhibit references, and defined-term usage. Contractual definitions are extracted first — the operative meaning of a clause depends on the definition of its key terms, and Clauseflint resolves these before flagging any position.

Cross-references between sections are mapped, so a limitation-of-liability carve-out in §12.4 that references an indemnification obligation in §9.2 is evaluated as a connected pair, not as isolated clauses.

Clause-category extraction runs across all sections. Categories include: change-of-control, MAC/MAE, indemnity, limitation of liability, assignment, non-compete and non-solicit, auto-renew and termination rights, data handling, audit rights, and governing law.

For each extracted clause, the engine records the section reference, the page and line range, the clause text verbatim, and the operative definition linkage where applicable.

Each extracted clause is assessed against the risk-threshold configuration set by the engaging team's counsel. Thresholds are configurable: a $50M deal might set a high-risk flag at indemnity caps below 3× fee; a smaller transaction might set 1× fee as the threshold.

Risk levels are three-tier: Flag (requires review), Watch (note for counsel), and Clear (standard market position confirmed). Clear designations do not appear in the flagged output but are recorded in the audit trail for completeness.

Flagged positions are compared against the market benchmarks embedded in the review configuration. These benchmarks are calibrated by contract type (NDA vs. MSA vs. acquisition agreement) and counterparty category (SaaS vendor, employment, IP licensor).

Benchmark data is drawn from the internal contract corpus and updated periodically. Clauseflint does not cite external databases by name — the benchmark is presented as a range ("market range: 2–5× fee for technology services MSAs") without attribution to a specific source.

The review output is delivered as a structured flag report and, for contract review engagements, as a Word redline. The flag report lists each flagged position with its section reference, risk level, and a brief description of the issue — written in clear legal English, not system language.

The redline uses standard Word track changes, with counsel-inserted comments corresponding to each flagged position. The audit trail export includes all Clear designations alongside Flags and Watches for a complete record.