The Challenge
A mid-market commercial real estate lender faced a critical bottleneck in their deal origination process. Every potential loan required comprehensive property research—searching comparable sales, extracting financial data, identifying broker contacts, and analyzing loan terms across dozens of properties.
Underwriters spent 4–6 hours per deal manually researching properties in CoStar. With 20–30 deals in the pipeline monthly, the team spent 80–120 hours per month on repetitive data gathering.
Manual lookups created cascading problems:
- Deals took 5–7 days from initial review to credit committee
- Human error in data entry led to incomplete comp sets
- Brokers went uncontacted because gathering their information was too time-consuming
- The team couldn’t scale without hiring more junior analysts
The Solution
We built an end-to-end agentic workflow system that automated the entire property research process using AI browser agents.
The system navigates CoStar like a human analyst would—but with perfect consistency and speed:
- Extracts property addresses from search results automatically
- Navigates to detail pages for each comparable property
- Extracts comprehensive loan data—origination amounts, interest rates, DSCR, maturity dates, prepayment terms, servicer information, and full financial statements
- Finds and extracts broker contact information from sales comp records
- Structures all data into a standardized format ready for analysis
Deployment took 3 weeks from kickoff to production.
The Results
- 37% reduction in deal cycle time — from 5–7 days to 3–4 days
- Research time cut from 4–6 hours to 15–20 minutes per deal
- 80–120 hours/month reclaimed from manual data gathering
- Comp sets expanded from 8–12 to 20–25 properties per deal
- 100% broker contact capture for available contacts
- 50% increase in deal capacity without adding headcount
Who This Applies To
Commercial real estate lenders where underwriters spend significant time on manual property research and data gathering. If your analysts spend more than an hour per deal on CoStar lookups, this addresses that bottleneck directly.