AI contract review — finding what the contract does not say
Written for in-house teams reviewing contracts at volume.
AI Tools Hub Editorial · published · updated
Checking for absence
An experienced reviewer spots a clause that reads wrong. What is hard is noticing a clause that is not present at all: liability caps, subcontracting rights, governing law and venue, data handling after termination. That is memory-matching work, and a checklist-driven system beats a human on the tenth contract of the day.
The risk of asking a general model
A general model returns plausible-looking comments. The problem is that its contract knowledge is largely drawn from US and UK practice.
- Advice that assumes a different legal framework
- Flags as unfavourable terms that are standard locally
- Omits terms that local practice requires
- Occasionally cites case law that does not exist
The practical sequence
Run the machine first to produce the list of omissions and flags, then have the human read with that list in hand. Reversed, the human reads twenty pages, becomes tired, and then has no appetite to evaluate the machine's comments.
Post-signature is the other half
Contracts auto-renewing because nobody noticed the date is a failure unrelated to review quality. A system that keeps the contract after signature closes that gap; two separate systems lose the thread at the moment of signing.
FAQ
Does it replace a lawyer?
No. First-pass review of routine agreements becomes internal work, which lets you choose which matters to send out. That is the gain.