Choosing a review strategy
Understand & Negotiate, High-priority only, or Strict — the review strategy sets how assertive the AI is and how much lands in your redline. Pick the one that matches where you are with the contract.
Updated 7 Aug 2026
The review strategy is the single most important choice in the review setup. It’s the last step — Approach — and it answers one plain question:
Where are you with this contract?
A first-look read of a contract you’ve never seen calls for a different emphasis than a final pre-signature check on a deal you’ve already negotiated twice. The strategy tunes two things to match:
- How assertive the review is — how hard it pushes, and what severity bar a finding has to clear to be worth raising.
- How much lands in your redline by default — the strategy sets a severity floor for what gets pre-included in the generated redline, before you’ve touched a single finding.
It does not change what the AI checks — that’s the rulebook’s job. The same rulebook clauses are evaluated every time; the strategy changes the posture the AI takes toward what it finds.
The three strategies
| Strategy | Where you are | What it does |
|---|---|---|
| Understand & Negotiate | Getting to grips with a contract — and likely to negotiate it | Surfaces everything so you understand the document, and flags what’s worth raising with suggested wording and fallbacks. Pre-includes medium-severity findings and up. The default. |
| High-priority only | You’re close to signing | Just the things that could genuinely cause problems. Pre-includes high-severity findings and up — the noise is filtered out. |
| Strict | You need every deviation on record | Flags every departure from the rulebook, however small. Pre-includes everything. For compliance passes and regulated templates. |
Understand & Negotiate (default)
“Understand the whole contract, with recommendations for negotiation — ideal for a first pass.”
The default, and the one most reviews want. It does two jobs in one pass:
- Understand the contract. It surfaces everything material — every risk, imbalance, unusual term, and gap — including things you may not be able to change, raised as awareness items so the picture is complete. This is the “what am I looking at?” read.
- Get ready to negotiate. For the issues where pushback is realistic, it gives you a defensible list of asks: suggested replacement language, and for the material ones an escalation ladder — the AI opens with the least-aggressive change that protects your position, gives a realistic fallback for each, and reserves hard, non-negotiable demands for genuine dealbreakers, so you can open cooperatively and escalate only if you need to. It also flags which asks are must-haves vs nice-to-haves vs optional.
Medium-severity findings and up are pre-included in the redline, so you land with a sensible starting markup you can trim rather than a blank one you have to build.
Understand & Negotiate vs Strict — both surface a lot, so what’s the difference? It’s the emphasis and how much pre-loads the redline. Understand & Negotiate is a working review: it explains the contract and turns the material issues into ready-to-use asks, pre-including medium-and-up. Strict is an enforcement pass: it flags every rulebook deviation, however trivial, and pre-includes all of it — for when you need a complete record of how a contract departs from your standard. One is for understanding and negotiating; the other is for holding the line.
High-priority only
“We’re close to signing — just the things that could genuinely cause problems.”
Late in the deal you don’t want to re-litigate the whole contract — you want to know if there’s anything left that could actually hurt you. This strategy raises the bar: only high-severity findings and up are pre-included in the redline, and the review’s emphasis is on dealbreakers and live exposure, not stylistic or nice-to-have improvements.
Strict
“Compliance pass — flag every deviation from the rulebook and redline all of it.”
For compliance reviews, regulated templates, and any situation where you need a complete record of how a contract departs from your standard. Every deviation is flagged and everything is pre-included in the redline. Expect more findings — that’s the point. You’ll typically work through them and switch off the ones you don’t want before generating the redline.
How the strategy shapes the redline
When a review finishes, each finding can end up in or out of the generated redline. Clment decides the starting position for each finding in this order:
- Your explicit redline action. If you’ve set a finding to Use this fix or No change, that always wins.
- Your wording. Choosing Custom instructions and saving your own wording for a finding includes it.
- Your call. Once you answer, it sets the default — Agree → include, Disagree → exclude.
- The strategy’s severity floor. For findings you haven’t touched yet, the strategy decides:
| Strategy | Findings pre-included in the redline |
|---|---|
| Understand & Negotiate | Medium severity and above |
| High-priority only | High severity and above |
| Strict | All severities |
The floor only ever applies to findings you haven’t engaged with. The moment you make a call, save custom instructions, or set the redline action yourself, your decision takes over. So the strategy gives you a sensible starting point — it never overrides a choice you’ve made.
This is why you can switch from a noisy Strict pass to a focused redline without losing work: decide the findings that matter, and the low-severity findings you ignored simply stay out.
The posture: a seasoned GC, not an over-eager junior
A frequent and fair worry about AI contract review is that it’s overzealous — that it invents problems, cites authority it can’t back up, hardens its stance when you ask it to “be tough”, and buries the real issues under a pile of trivia. Clment reviews are built to behave like a seasoned general counsel instead, and this posture holds on every strategy:
- Stay grounded. The AI cites only what the contract or your rulebook actually supports. It won’t invoke statutes, regulations, cases, or standards it can’t tie back to the document or the rulebook. If it thinks something external applies but can’t ground it from what it was given, it marks the finding needs verification and says so — rather than asserting unverifiable authority as fact. (See the source badges in Understanding finding verdicts.)
- Hold a reasoned line. The AI won’t inflate severity or harden its position just because your instructions are framed combatively — and it won’t drop or soften a genuine issue because someone asserts the opposite without a substantive reason. It changes its assessment only on the merits.
- Be measured, not maximal. It prefers the least-aggressive position that protects you, and reserves critical severity for genuine dealbreakers.
The strategy you pick changes how much the AI surfaces and redlines — but it never licenses the AI to overreach. Even Strict flags small deviations without inflating what they mean.
Add an executive summary
Below the strategy picker is an Add an executive summary toggle. It’s independent of the strategy: tick it on any review and you’ll get a short, plain-English overview written for a non-lawyer audience — alongside the detailed findings, not instead of them. Useful when the review is going to a deal owner, a founder, or a colleague who needs the gist without reading every finding.
Which should I pick?
- Getting to grips with a contract, or going back to the other side? → Understand & Negotiate. The default — it explains the contract and hands you ready-to-use asks in one pass, so it’s the right call whether you’re reading it for the first time or about to push back.
- Final check before signature? → High-priority only. Surface what could still hurt you, skip the noise.
- Compliance / regulated template / need a full record? → Strict.
You can re-run a review with a different strategy as many times as you like — though a re-run is a full review, so switch strategy when the posture genuinely no longer fits (e.g. moving from active negotiation to a final high-priority-only check before signature), not out of habit.
See also
- Reviewing contracts — the four-step setup this is the last step of.
- Understanding finding verdicts — the four controls you use on each finding, and the source badges that show what a finding is grounded in.
- Generating redlines — how the strategy’s severity floor feeds the redline.
- Creating a rulebook — what the AI checks (vs how assertively), with a built-in strength check on your instructions.