Finance

Every acquisition was underwritten by hand, one model at a time, from whatever financials the current owner kept. Their own AI now builds the model from those financials, reads the books whichever way an owner writes them, and checks its own work before it shows a number.

The numbers.

Analyst seat retiredSalary line removed, the work retained
$75,000/year
Time savedPer proforma, on every proforma
3 hours/analysis
Assets modeled2.16 million square feet across 13 states
25
Calibration match25,723 of 25,723 computed cells matched the company's own model
100%
Account mapFive different chart of account dialects read into one model
695 rows
Validation gatesAll must pass before a number is released
5
Data points heldAcross every deal built, all queryable
149,316
Assumptions recordedEach one written down and answerable later
514

What changed.

Three hours came off each analysis, measured by the client on every model they ran. Twenty five assets have been modeled this way, 2.16 million square feet across 13 states. Checked against the company's own model, all 25,723 computed figures matched. Every assumption it makes is written down and can be asked about later.

How this was published.

The work and the figures below, drawn from the client's live system in September 2026, published with the company name withheld at the client's request.

The client agreed to this study in writing and asked that their name be left out. We do not print the approver either, because naming them would identify the company.

Published September 2026.

This site sets no tracking cookies. Optional tracking stays off unless you enable it.