Check four things when you use AI for multifamily underwriting: how market rent is supported, how concessions are treated, whether the unit mix stays visible, and where each expense line came from. Every tool demos well on a clean rent roll. What arrives is an export with merged header cells, unit types labeled three ways, concessions in a notes column and a total that does not quite tie to the T-12.
Prepared by Titleman AI, on behalf of the Titleman team. Figures are from finished Titleman runs and measurements published on the pages linked here, made on Titleman QA accounts or read from production, not from client files. A run is a draft for a professional to check, not an appraisal, investment advice or a credit decision.
1. Loss to lease
Loss to lease is the gap between market rent and in-place rent. A large gap can be real upside as leases roll. It can equally mean whoever set the market rent was optimistic, in which case the gap is a forecast presented as a fact. Ask: what is market rent based on, and which leases support it?
2. Concessions
One month free on a twelve-month lease is about an 8% rent reduction. If the face rent goes into the model and the concession is treated as one-time, income is overstated, then multiplied, because value is a function of income. Put a rent roll with concessions in a notes column in front of the tool and see whether effective rent comes out reduced.
3. Unit mix
A blended average rent hides the mix. Full one-bedrooms at strong rents and vacant three-bedrooms can show a healthy blend and a broken income story. Ask for rent, occupancy and lease expiry by unit type, with counts.
4. Expenses
Borrowed benchmarks do the most damage here. Property taxes reassess on sale in many jurisdictions and not in others, insurance has moved unevenly by geography, and payroll differs by property size. "Market standard" is not a source. Acceptable answers are the property's own T-12, a named comparable or a stated assumption you can argue with.
Compare the working, not the conclusion
Take one property you underwrote yourself and run it. Then ask:
- Is every figure traceable to a rent roll cell, T-12 line or comparable?
- Does the tool flag units it could not parse, or drop them silently?
- Where evidence is thin, does it state a range and why, or a confident point?
- Can you change one assumption and have everything dependent move consistently?
Also test the documents against each other. On a sample 120-unit package the memorandum said 96.0% occupied; the rent roll showed 110 of 120 units, 91.7%, and the memorandum's expenses of USD 678,555 were USD 283,311 below what the T-12 showed.
What it does not replace
None of these checks replaces a licensed appraisal, physical diligence on the units behind the rent roll, estoppels, or the judgment of whoever prices the deal.
See how to read a rent roll and the failure modes in full. To test Titleman, send a rent roll you know.