At the point at which (1) you are doing something that takes a written contract, and (2) you identify that it is beyond your confidence of what you can deal with on your own, why would you trust an AI agent that can't, for example (based on my testing), consistently remember which party to a contract it is advising, or which on the same contract will refer to the same provision, on different runs, as both admirably clear as a positive or dangerously vague as a negative?
At the point at which you need a second eye on a contract, you probably want to have some confidence that it is better than a magic 8-ball.
Compared to the cost of a mistake of a potentially hallucinating robot? “The LLM was wrong” will not fix a breach of contract or similar legal mistake. The tools make an excellent job aid like an IDE assistant, but human review is still mandatory.
It's not that expensive, though. If you're regularly dealing with contracts complex enough that a solution like this has appeal, you probably also have a decent income and a couple hundred dollars won't hurt much. Missing, or misunderstanding, something in the contract can also cost a lot more than what an attorney would charge you to review the contract.
But to each their own. For me, using something like this is riskier than I'd be comfortable with.
At the point at which you need a second eye on a contract, you probably want to have some confidence that it is better than a magic 8-ball.