Contract review work lately keeps surfacing the same observation: AI has gotten genuinely good at producing a clean first draft. Not because it understands the law better than a lawyer, and not because it grasps commercial intent better than an experienced contract manager — it’s simply fast. Feed it a term sheet and it can return a structured agreement in minutes. Give it two versions of the same document and it will catch small wording changes a tired eye might miss on the fifth read-through of the day.
That speed is genuinely useful, and it shows up in a handful of specific places: turning rough terms into something usable quickly, flagging clauses that are commonly present in a given contract type but missing here, converting dense legal language into something a business team can actually act on, and catching inconsistencies that build up across a long negotiation history.
Where it falls short is less about the words on the page and more about everything behind them. A model doesn’t know how much risk a particular organization is willing to carry into a project. It doesn’t know why a clause was worded a certain way three renegotiations ago, or which counterparty pushed for it and why. It can’t tell which sentence — one that reads as boilerplate to anyone skimming it — is the one that ends up costing millions if a dispute ever goes to arbitration. That kind of judgment comes from sitting across the table during enough negotiations to recognize the pattern before it becomes a problem.
This isn’t an argument against using these tools. It’s closer to a reframing of where the value actually sits. Producing a draft is becoming the easy part of the job. The harder part — deciding what risk to accept, where to push back, and which battles are worth the negotiating capital — was never really about the writing in the first place. It was about judgment built from doing this work long enough to recognize the shape of a bad clause before it’s fully written out.
For contractors and project owners, the practical takeaway is straightforward: drafting speed is not the same as risk protection. A polished-looking contract can still carry exposure that no AI tool flagged, because the tool was never looking for it. Independent review — someone reading with the specific risk tolerance of the project in mind — still matters, arguably more than before, now that the volume of paperwork moving through a project has gone up while the time spent scrutinizing any single clause has gone down.
The contracts getting written are only going to get faster to produce. The ones getting negotiated well will still depend on people who have seen enough of them to know where the real risk is hiding.