The workflow before
When a landlord and tenant agreed on a proposal, attorneys drafted the lease from it clause by clause, then traded redlines over whether each clause was market standard. A first draft took days and carried legal fees on both sides. MarkIT CRE wanted users to upload a proposal and get a market-norm draft back immediately, tilted toward the landlord or the tenant as they chose.
What we built
We built a lease drafting API grounded in a curated library of real lease clauses. Retrieval pulls the closest market-norm language for each section from a vector database, so the model drafts from proven clauses instead of inventing them.
A user uploads a proposal, sets favorability toward landlord or tenant, and gets a full draft lease back in one call. The API also edits surgically: it rewrites a single clause without regenerating the whole document, which is how redlining actually works.
We wrote separate prompt sets for each task a negotiator needs: summarize a clause, explain the differences between two versions, assess which side it favors, and rewrite it to market norm. The MVP shipped as a Docker container ready to deploy on AWS.
Before and after
| Measure | Before | After |
|---|---|---|
| First draft from proposal | 2 to 3 days of attorney drafting | Minutes, one upload |
| Single clause revision | Manual redraft and redline | Targeted rewrite through the API |
| Market-norm check | Attorney judgment, clause by clause | Automatic comparison against a standard lease corpus |
What's next
The MVP proves the drafting and redlining loop. The next step is production deployment on AWS and an L3 negotiation workflow, where the system proposes counter-language on each redline and a person approves the exceptions.
