Lease abstraction at 10× speed, with full audit trail.

Scope

Stand up an AI-assisted lease intake and abstraction workflow that ingests incoming PDFs, extracts standard fields and key clauses, validates against business rules, and pushes structured records into the existing property management system — keeping a human in the loop for every exception.

• Every new lease and amendment required a manual abstraction pass. The backlog regularly sat at 2–3 weeks, blocking accounting and asset management downstream.
• Critical dates — options, escalations, expirations — were being missed because they were buried in PDFs no one had time to re-read.
• Each abstractor formatted things slightly differently, which meant downstream systems had to be normalized by hand before reports could be trusted.

• Mapped 40 representative leases — retail, office, industrial — and cataloged every field the team needed extracted, plus the business rules that decided what 'normal' looked like.
• Designed a workflow that splits work into three lanes: clean leases auto-extracted and queued for one-click approval, normal-but-flagged leases routed to a junior reviewer, and exceptions escalated to senior staff with the specific clause highlighted.
• Built directly against historical abstractions as the gold-standard test set. Iterated until extraction accuracy matched the senior abstractor on every required field.
• Deployed in production with a feedback loop — every reviewer correction trains the next batch's extraction rules.

Measured after the first 90 days.