Understanding finding verdicts

The four controls on every AI review finding — verdict, comment, include-in-redline, redline instruction. What they mean and how to use them well.

Updated 17 Jun 2026

Every AI review produces a set of findings — one per clause your playbook told Clment to look at. Each finding has four independent controls. They’re independent on purpose: the AI’s analysis, your reasoning, your verdict, and what ends up in the redline are separate questions.

The four controls

1. Verdict — what you think of the AI’s recommendation

Three values:

  • Agree — you agree with the AI’s recommendation for this clause.
  • Disagree — you don’t agree.
  • Partial — you agree with part of it; you’ll explain in the comment.

The framing trips people up: “Agree” means agree with the recommendation, NOT “agree with the clause as drafted”. If the AI recommends rewording a liability cap and you say Agree, you’re saying yes, reword it.

We chose this framing deliberately because it aligns the verdict with the action. “Do I agree?” maps directly to “is this change going in the redline?”. The alternative — “do you agree with the clause?” — is ambiguous because the clause might already be fine and the AI flagged it for a different reason.

2. Comment — your reasoning

A free-text note explaining your reasoning. Required when you pick Partial, optional otherwise.

Comments stay attached to the finding and show up in:

  • The exported review report.
  • The activity log for the contract.
  • Anyone else’s view of the finding (reviews can be shared with colleagues).
  • The audit log for compliance reviews.

Useful comments capture why you decided what you decided — not just what you decided. Future you (or your replacement) will thank present you for “Partial — agreed indemnity scope needs widening but their proposed language is too broad; counter with our standard cap” instead of just “see redline”.

3. Include in redline — should this end up in the document?

A boolean: should this finding’s change end up in the generated redline document?

Clment decides each finding’s starting position in a clear order of precedence — the first rule that applies wins:

  1. Your explicit toggle. If you’ve flipped include on or off, that always wins.
  2. A redline instruction. If you’ve written replacement wording (control #4), it’s included.
  3. Your verdict. Once you’ve given one — Agree → include, Disagree → exclude, Partial → you tell us.
  4. The review strategy’s severity floor. For findings you haven’t touched yet, the review strategy you picked decides: Explore includes nothing, Negotiation includes medium-severity and up, High-priority only includes high-severity and up, and Strict includes everything.

So a finding you’ve engaged with follows your decision; a finding you’ve ignored follows the strategy. This is what lets you run a noisy Strict pass and still produce a focused redline — decide the ones that matter and the rest fall away.

You can override at any time. Why have this as a separate control rather than just inferring from the verdict?

  • You might Agree with the analysis but choose not to redline it now — for example, you’re saving the redline for a later round, or this finding is informational (“the indemnity scope is broad — flagging for awareness, not changing”).
  • You might Disagree with the AI’s recommendation but still want a redline edit — your edit, not theirs. The redline instruction (control #4) is how.

4. Redline instruction — your wording, not the AI’s

Optional free-text override of the AI’s suggested replacement language. If you want the redline to use your wording instead of the AI’s, write it here. The redline generator uses your text verbatim if present, the AI’s otherwise.

Useful when:

  • The AI’s wording is too aggressive (or too soft) for the relationship.
  • You have house-style language for this clause that needs to be used verbatim.
  • You’re negotiating in a specific direction — e.g. “the AI recommended uncapped indemnity, but we’d accept a $5M cap” — and you want the redline to reflect the actual position you’ll put forward.

A worked example

Finding: “Limitation of liability cap is set at fees paid in the prior 12 months. Recommended: increase to fees paid in prior 24 months.”

ScenarioVerdictCommentInclude?Redline instruction
Accept the AI’s position as-isAgree(empty or short note)✓ Yes(empty — use AI’s wording)
Agree with the principle but use our standard capAgree”We want 24mo cap as standard”✓ Yes”Cap = fees paid in prior 24 months; carve-outs for breach of confidentiality, IP infringement, gross negligence.”
Cap is fine, no change neededDisagree”Standard relationship — 12mo cap is acceptable”✗ No(empty)
Push for 36mo, not 24moPartial”Agree with principle but pushing for 36mo on this risk profile”✓ Yes”Cap = fees paid in prior 36 months …”

Sign-off

When you’re done working through the findings, Sign off locks the review. The review status flips to signed off and a snapshot of the finding states is preserved. Future edits create a new revision (REV-13, REV-14, …) so the original is preserved as the record-of-decision.

You can assign a review to a colleague before sign-off (e.g. for a partner review). They get an in-app notification + an email with the ?review=REV-12 deep-link.

See also

Still have questions?

Instant article search