From Sticky Notes to Smart Agents: How AI Is Reshaping Contract Review

In 2019, contract review meant hours with a highlighter and sticky notes. Most of the time wasn’t judgment, it was finding where terms might live across thousands of pages.

Machine learning OCR with Human-In-The-Loop promised to revolutionize review, but at ten cents a page it was expensive.

When GPT-3 arrived in 2020, I experimented. It couldn’t be trusted for legal interpretation. Hallucinations often flipped contract clauses to mean the exact opposite of reality, and GPT-3 would make up precedent, a bad precedent itself. However, it could reliably index unlabeled conditions. That was enough to cut hours into minutes: I still made legal interpretations, but AI gave me a fast map of where to look.

What OpenAI describes in their Contract Data Agent is the next step. Not a lawyer, not an intern, but an always on paralegal. Instead of ad hoc searching, it can extract structured data across thousands of contracts, output it consistently, and feed it into workflows. Humans remain in the loop for interpretation, but the drudgery of “find and tag” is quick and scalable. I would not need two interns and a week to get through five bankers’ boxes of paper.

There’s a lot of focus on the consumer side of AI. The messy, sometimes absurd errors it produces. Here’s an excellent example of where AI can directly address knowledge industry expert’s workflow. Progress in AI tools isn’t about replacing judgment (yet). It can remove serious bottlenecks from labor hours. From sticky notes, to search helpers, to structured extraction. the work is still ours, but the grind is less.

Again, for now.