What the work involves
You work through queues of contract-centered tasks that test whether a model can behave like a competent transactional lawyer. A typical session might mean drafting a limitation-of-liability clause against a specified risk allocation, comparing two model redlines of an indemnity provision and explaining which one a client could actually sign, or writing a prompt that exposes a model's tendency to hallucinate a survival period that the agreement never granted. Other tasks ask you to produce the reference answer first — the issues list a reviewing associate would flag in an MSA, a clean summary of assignment and change-of-control restrictions, a defensible position on an ambiguous notice provision — and then score model output against it.
Ratings are not thumbs up or down. Each judgment is paired with written reasoning that a reviewer without your background can follow: which term is wrong, why it matters commercially, what the correct treatment would be. Reviewers weight that reasoning as heavily as the score itself, and vague justifications get tasks sent back.
What the platform screens for
- Verifiable transactional experience — the agreement types you've drafted and negotiated, on which side, at what deal size
- Depth that survives follow-up: the AI screener will push on a clause you mention and ask how you'd handle a specific counterparty pushback
- Whether you can distinguish a genuinely wrong contractual statement from a stylistic preference or a defensible alternative approach
- Clear written English under time pressure, since nearly all output is prose
Logistics
Fully remote and asynchronous. You claim tasks when you have time; there are no standing meetings and no assigned shifts. Most contributors treat this as 8–15 hours a week alongside practice, though some campaigns open larger blocks with short deadlines. Expect a calibration period at the start, where early submissions are reviewed closely and pay may be tied to a shorter onboarding batch before you move to the full rate. Work is contractor-based, per-task, with no volume guarantee — campaigns start and stop with client demand.