Scaling legal review
How to scale legal review without hiring more lawyers.
Legal review does not scale by adding reviewers, because the thing that takes the time is judgement and judgement lives in a small number of heads. It scales when that judgement moves out of the heads and into a playbook that runs on every contract before a person sees it.
Why legal review does not scale by adding reviewers
A commercial legal team of four handles a few hundred contracts a year comfortably. At a thousand it is drowning, and the obvious fix is a fifth lawyer. It rarely works, for three reasons.
The first is ramp. A new reviewer is not productive on your paper for months, because your standard is not written down anywhere they can read it. They learn it by having their work corrected, which costs the senior reviewer time in the same period you were trying to buy time back.
The second is drift. Every reviewer you add is another interpretation of the same clause. Five people applying an unwritten standard produce five standards, and the difference only surfaces when a counterparty points at two of your contracts.
The third is that most of the work was never worth a lawyer. In a typical queue, the majority of contracts are entirely standard and fail nothing. Somebody senior read them anyway, because until you read it you do not know which kind it is.
The four ways teams try to scale legal review
| Approach | What it fixes | Where it runs out |
|---|---|---|
| Hire more reviewers | Raw throughput, eventually | Cost is linear, ramp is slow, and the standard fragments as the team grows |
| Publish templates | What a contract should say when you draft it | Says nothing about counterparty paper, and nothing stops a template being edited |
| Write a review checklist | Consistency, on the days people use it | It is advisory. It is not applied to every contract, and nobody can prove it was applied to any |
| Add AI summarisation | Reading time | Produces an opinion the reviewer still has to check, so the decision stays with the person |
| Publish a playbook that runs | The judgement itself, applied to every contract before a person opens it | Genuine exceptions still need a lawyer — which is the point |
The first four make review faster. Only the last one removes contracts from the queue.
Move the judgement out of the review
The senior reviewer is not slow. They are doing something that has never been written down: holding a set of positions about liability, indemnity, term, notice, governing law and data protection, and comparing each contract against them. That set is the asset. It is also the bottleneck, because it exists in one place.
Writing it down changes the shape of the work. Once the positions are rules, they can be applied to every contract automatically, in the same way, before anyone reads a page. Review stops being read this and form a view and becomes three rules failed, here is the text that failed them, decide.
The measure worth tracking. Not average review time. Track the share of contracts that reach a human at all. A team that reviews 1,000 contracts and only opens 180 of them has scaled. A team that reviews 1,000 contracts twice as fast has bought itself a year.
What a scaled review process looks like
The standard is written down and published
Your positions become a playbook: rules with thresholds, and the approved language behind each one. It has a version and a publication date, so there is always an answer to "what was our standard in March?"
Every contract is checked before a person sees it
Contract analysis evaluates the whole active playbook against the document in under a minute — counterparty paper and your own drafts through the same pipeline.
Clean contracts leave the queue
A contract that fails nothing does not need a reviewer. It carries a record showing which playbook version cleared it and what was checked, which is more evidence than a silent approval ever produced.
The lawyer reads the exceptions
What reaches a person is a short list: the rules that failed, the extracted value that failed each one and the passage it came from. That is a five-minute job on a document that used to take forty.
The business drafts inside the rules
Sales generates its own NDAs and order forms from the clause library, so routine paper never enters the queue in the first place. This is the part that changes the volume rather than the speed — see sales contract compliance.
Exceptions feed back into the standard
When the same rule is overridden repeatedly, the rule is wrong. You change it once, publish a new version, and the correction applies to every future contract instead of being re-learned by each reviewer.
How to get there in a week
This is not a transformation programme. The work is mostly deciding what your positions actually are, which is a conversation your team can have in an afternoon.
- Day one. Upload your templates and thirty signed contracts. Lexnus drafts a playbook from the positions you have been taking in practice.
- Day two. Read the draft with whoever holds the standard today. Most of the argument happens here, and it is the useful kind — the rules make disagreements explicit that were previously invisible.
- Day three. Run the published playbook against last quarter's contracts. The pass rate tells you whether the rules are calibrated. Too many failures means a threshold is wrong, not that your paper is bad.
- The rest of the week. Point live traffic at it, starting with one contract type. NDAs are the usual first choice: high volume, low variation, and nobody enjoys reviewing them.
If most of your volume is inbound paper on someone else's template, start with counterparty review instead.
Questions
Does this replace legal review?
No. It removes the contracts that never needed a lawyer and puts the evidence in front of the lawyer for the ones that do. The judgement is still legal's — it is just made once, when the rule is written, instead of once per contract.
What share of contracts stops needing review?
It depends entirely on your paper and how strict your rules are. The useful exercise is to run a published playbook against last quarter's contracts and count. That number is specific to you, and you can have it on day three.
Our standard genuinely varies by deal size and region.
That is a rule, not an exception to rules. Thresholds can differ by contract type, counterparty and jurisdiction. What the playbook removes is unintentional variation — the same deal reviewed twice and treated differently.
Who maintains the playbook?
Legal, in the same way legal maintains templates today. Changes are versioned and published, so the standard has an owner and a history rather than living in an inbox thread.
What about contracts drafted with AI?
Same pipeline. An agent drafting through MCP integration is checked against the same published rules as a person, which matters more as more of the first draft comes from a model. See AI contract drafting.
Find out how much of your queue is already clean.
Upload last quarter's contracts. Lexnus drafts a playbook from your own paper, runs it across all of them and shows you how many would have reached a lawyer. Free trial, no card.