How to redline a Word document

• Deal teams circulate a draft where comments duplicate, contradict, or bury the real asks.
• Track Changes noise (fonts, spacing) hides a material definition or obligation shift.
• You are about to run structured AI review and need the manuscript stable enough for consistent spans.

• Drive-by comments without a proposed edit—fine for questions, expensive for execution.
• Global style “fixes” that create hundreds of low-value changes before substantive review.
• Implicit rewrites in comment text that never become tracked insertions—easy to miss at signing.

1. Freeze scope: decide what is in-round (definitions, economics, reps) vs out-of-round (formatting polish).
2. Normalize the baseline: one heading hierarchy, consistent list numbering, stable styles—so later diffs map to intent.
3. Use comments for questions, inserts for proposals: keep “why” in comments and “what changes” in tracked text where possible.
4. Tag severity in your team’s language (blocking / material / cosmetic) so downstream readers triage faster.

After the manuscript is stable, DraftLens can run structured reviewers on the same file and surface a merged issue ledger—useful when you want machine-assisted triage before partner markup. Review mode keeps judgment-first outputs; do-not-change locks carry your “must not drift” phrases into model context.

• Negotiation strategy, privilege decisions, and regulatory interpretation.
• Counterparty-facing tone and relationship risk—still human-led.
• Final sign-off: models can miss context; treat outputs as inputs to your existing QC.

• Every blocking comment has an owner and a proposed resolution path.
• Defined terms and party names match your control list (locks help here).
• Cross-references resolve after insertions (schedules, exhibits, section xrefs).
• You have a second human read on anything labeled material—even if AI agreed.