A general chat assistant can do arithmetic on the figures you paste into it, but underwriting a deal is more than arithmetic: it means finding the records behind the property, tying the documents to each other, recomputing the model and saying what could not be checked. The fair test is a deal you already closed. Run it, then compare the working, not only the answer, with your own file.
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.
How a good formula gives a wrong answer
The clearest case is our own. In an underwriting of a 416-unit apartment property in Tampa, the first draft returned a levered IRR of 53%, an equity multiple of 6.36x and a DSCR of 2.98x. The formulas connecting price, debt, income and exit value were all correct. One input was wrong: the purchase price cell held the assessed value, USD 35.21 million, a figure used to compute a tax bill, not a price. The corrected run derives price from Year 1 income at a stated cap rate, passed four automated checks and returned an IRR of 8.66%.
Arithmetic on a wrong input is still wrong, and a flattering error does not look like one. The full case is on why runs disagree.
Five tests for any assistant
| Test | What a pass looks like |
|---|---|
| Reads the documents behind the OM | The OM's NOI and occupancy are checked against the T-12 and the rent roll |
| Recomputes formulas | A model cell saved at USD 15,120,000 is caught when its own formula gives USD 14,040,000 |
| Places every claim | Each flag points to a page or cell |
| Counts what it checked | Items found and items checked are both stated: 1,453 and 1,453 in one run |
| Lists what it could not check | Six items in that run, such as no address and no lease files |
The first two rows come from one Titleman run on a sample 120-unit package, described on how to analyze an OM with AI. The OM stated NOI of USD 1,312,400; its own T-12 totaled USD 1,029,089.
Where Titleman fits
Titleman is built to be checked this way: it returns the answer with each figure tied to a record or a cell, and lists open items. It is still a draft. A named professional on your team signs the output.
What it does not replace
No AI tool, ours included, replaces a licensed appraisal, a title search, a site inspection, legal review of the loan or lease documents, or the committee's own judgment.
To compare the two approaches on a deal you know, request a demo and send the package.