Case study: real estate underwriting
Pre-offer underwriting in about 45 minutes, not two days
An AI underwriting system built for E&S Properties and Pencil This Underwriting, the real estate businesses Stoker AI grew out of. It finds deals, reads the documents, asks the broker what's missing and produces a full underwriting report before anyone makes an offer.
Measured across several versions of the system: 250 properties for E&S Properties over 18 months and 100 for Pencil This Underwriting's clients over the last 6. Hours saved compare about two 8-hour days of manual underwriting at the same depth with about 45 minutes per property.
The clients
Two businesses, one engine.
E&S Properties buys and operates value-add multifamily property. Pencil This Underwriting analyzes deals for other investors, for a fee.
Both had the same problem. Underwriting a deal properly before making an offer means a professional working two to four days on one property. That's expensive, so most deals get a rough look, offers get made on thin numbers, and the real underwriting happens after the letter of intent. When the numbers change, the deal gets retraded or dies.
For small deals and small operators, that depth was simply out of reach. Nobody pays an underwriter for three days to look at a 15-unit building they may never bid on.
What we built
From listing to letter of intent.
The system runs on its own throughout the day. People step in at the two points that matter: what gets asked and what gets offered.
- 01
Find
Every night the system searches broker listing sites and broker websites in the states and markets the buyer targets, and pulls in anything new.
- 02
File
Deal packages that arrive by email, and packages downloaded from broker sites after an NDA, are read, sorted and filed to the right deal automatically: rent roll, T-12, offering memorandum, appraisal.
- 03
Screen
Each deal runs through the buy box: property class, size, population growth, rent trend, title restrictions such as rent caps, occupancy and price ceiling. Most deals stop here.
- 04
Ask
For deals that pass, the system lists what the package left out and drafts the questions for the broker. A person approves them before they go. Replies are matched back to the right deal from the inbox.
- 05
Underwrite
Full pre-offer underwriting runs: rents, expenses, capital needs, renovation plan, exit value, financing and returns, ending in a maximum allowable offer.
- 06
Offer
If the maximum offer lands within the buyer's tolerance, a letter of intent is drafted on standard terms. A person reviews it and decides whether to send it.
Inside the underwriting
The depth usually saved for due diligence.
A typical report runs about 20 pages and covers what a professional underwriter would, before any offer is made.
- Rents from evidence. The property's own recent leases and finish-matched comparables, not the seller's "market rent" column.
- Tier and tail. Where each unit's rent sits against its market, so a building whose long-time tenants pay above market isn't mistaken for upside.
- Expenses normalized. Taxes reset to the buyer's price, insurance at a real quote, payroll and repairs at realistic levels.
- Seller data checked. Every figure the seller supplied is compared against the documents, and every mismatch is flagged.
- Capital needs. Building systems and roofs the offering memorandum leaves out, priced from local quotes where available.
- Renovation tier picked by return. Four renovation levels, each costed with local labor rates and local material prices, and each tested for whether the rent lift pays for the work. The goal is to improve to what each market will pay for, not beyond it.
- Exit value built in layers. A market cap rate adjusted for class, submarket, size and permanent features, with the reasoning written out.
- Financing and returns. The buyer's own financing, return hurdles and hold period, ending in one maximum allowable offer.
- What moves the number. Sensitivities on every key assumption, a risk list and negotiation points.
For Pencil This clients, the same engine runs on each client's own financing plan and goals, so two investors can get two different maximum offers on the same building, both correct for them.
How the work is split
AI reads. Rules count. People decide.
AI does the reading and research
Offering memorandums, rent rolls and T-12s in every format, broker emails, listing pages, comparable rents and market data.
Rules do the math
Every rent, expense, return and offer price is calculated the same way every time, and every number traces back to its source.
People make the calls
Questions to brokers and letters of intent are drafted for approval. Nothing is sent and no offer is made without a person signing off.
Results
Fewer deals, better ones.
E&S Properties went from analyzing a few deals by hand to screening 20 to 30 properties a day.
Only a handful become letters of intent. In practice, about one deal in 50 to 75 makes sense once it's underwritten this closely, which is the point: the system filters out the noise so the time goes to the deals worth pursuing.
Because the numbers are solid before the offer, deals rarely need to be retraded. A price changes only when the site visit shows something the documents didn't.
And small deals now get the same depth of analysis as large ones, which was never affordable when it took a person days.
For your business
The same pattern fits most back offices.
Messy documents in, rules applied the same way every time, a person approving what goes out. That's reconciliation, billing, reporting and intake in most businesses, not just real estate.
If your team rebuilds the same analysis or report by hand every week, we can usually automate most of it. See how we work, or see what we do for real estate.
Want this kind of system for your business?
Tell us how the work gets done today. A short call or questionnaire, then a written plan and a fixed price.
Free, by phone, no obligation. We call you at the time you pick.