Litigation review is the most expensive reading in the building. On a large matter, a team of contract reviewers works through hundreds of thousands of documents, one at a time, deciding what is responsive and what is privileged. It is slow, it is costly, and most of it gets billed back to the company by the gigabyte.
That is the work we built Discovery AI to do first.
Where does litigation review get expensive?
The cost is not the filing or the strategy. It is the reading. Before a case can be argued, someone has to look at every document and decide what is relevant and what is protected. On a mid-size matter that can mean weeks of reviewer time, and the bill scales with the volume of data, not the value of the case.
Two decisions carry the most risk. Responsiveness, because a missed document can mean a missed obligation. And privilege, because a document handed over by mistake can waive a protection you cannot get back. Both are judgment calls made hundreds of thousands of times, under time pressure, by people reading fast.
What does the AI do, and what does your team still own?
Our agents read every document and make the first decision on responsiveness, issues, and privilege. They do not skim and they do not sample. They read each one and return a call with a reason attached, so a reviewer is checking a recommendation instead of starting from a blank page.
What the agent does
- Reads every document, not a sample
- Makes the first call on responsiveness, issues and privilege
- Attaches a reason, cited to the text
What your team does
- Confirms or overrides each call
- Works the close calls first
- Owns the strategy and the outcome
Your lawyers still own every decision that matters. The agent takes the first pass off their plate so their time goes to the close calls, the strategy, and the documents that actually move the case.
How do you defend an AI-assisted review?
Every agent decision cites the source text it relied on and is logged with the actor, the timestamp, and the reasoning. When opposing counsel or a court asks how the review was done, the record is already written. You can validate the results with a statistical sample and show the work, rather than asserting that a percentage was eyeballed.
We cover what that takes in What “defensible AI review” actually means.
Why price per matter instead of per gigabyte?
The old model charges by the gigabyte, so a routine matter with a large mailbox costs the same as a bet-the-company case. We price per matter, so your finance team can plan the line in advance. The villain here is not the law firm. It is the per-gigabyte cost model that makes every matter more expensive than it needs to be.
The numbers behind that are in The economics of bringing litigation document review in-house.
Does the same approach work beyond litigation?
Once the engine reads and reasons over documents, the same capability applies to the work next door. Internal investigations, subject access requests, and confidential-information checks are the same job underneath. We lead with litigation because it is the hardest and the one we know best. The rest follows.
Where to start
If you run litigation review in-house, book a demo.