This guide is general information, not legal, tax or investment advice. Your operating agreement, offering documents, counsel and CPA govern your deal.
On this page
What AI underwriting tools actually do
Underwriting a commercial real estate deal starts with retyping. A broker sends an offering memorandum (OM), a rent roll and a trailing twelve-month operating statement (T-12), usually as PDFs, sometimes as scans, and an analyst keys the numbers into a model before any thinking happens. AI underwriting tools go after that step.
The products differ a lot, but most of them do some mix of the same six jobs:
- Document parsing. Read the rent roll unit by unit, the T-12 line by line and the OM for property facts, from PDF or Excel, with OCR for scanned pages.
- Classification. Map each operating-statement line to a standard chart of accounts, so "R&M - Turnover" and "Make Ready" land in the same bucket and the expense ratio means something.
- Comps and market data. Pull rent comps, sale comps, expense benchmarks or tax and insurance estimates, from a vendor dataset, a licensed provider or public sources.
- Model population. Fill a cash-flow model: the vendor’s own, a template you configure, or your existing Excel model.
- Screening and summaries. Produce a tear sheet, flags against your buy box, or a first-pass memo.
- Questions on the deal. Let you ask about the documents in plain English and get an answer that points back to a page or a row.
Where the errors hide
Parsing a clean, digital rent roll is the easy case. The output goes wrong in predictable places, and a small error in the wrong place moves the value of the deal. Vendors publish accuracy figures, and they are the vendors’ own measurements on their own test sets. Your documents are not their test set.
These are the checks worth building into every review, whichever tool you use:
| Input | What tends to go wrong | How to check it |
|---|---|---|
| Rent roll: units | Units dropped across page breaks; model, down or employee units counted as occupied | Unit count and square footage tie to the OM and the unit mix |
| Rent roll: rents | Market rent read as contract rent; concessions, loss to lease and month-to-month leases missed | Sum of contract rent ties to the rent roll total; spot-check ten units |
| T-12: totals | A month missing or double-counted; subtotals parsed as line items | The twelve months add up to the statement’s own annual total |
| T-12: classification | One-off items (insurance proceeds, a lawsuit, a roof) treated as recurring | Read every line above a materiality threshold you set |
| T-12: normalising | Seller’s management fee, payroll or utilities carried over unchanged | Replace with your own costs; reassess property taxes at your purchase price |
| OM figures | Broker pro forma numbers read as actuals | Every OM figure in the model is labelled pro forma or actual |
| Comps | Old, distant or different-vintage comps carried into rent growth | Check date, distance, year built and unit mix of each comp you keep |
Why one misread line matters
A fictional example. A 120-unit property’s T-12 shows $1,020,000 of net operating income. In month seven there is a $60,000 insurance recovery after a burst pipe, and the parser classifies it as other income. The model now treats it as recurring.
At a 6.0% cap rate, $60,000 of NOI is worth $1,000,000 of value ($60,000 ÷ 0.06). The real trailing NOI is $960,000, so the price your model supports is about $1,000,000 too high, before debt sizing, which also keys off NOI, makes it worse. Nothing about the parse looked wrong; one line sat in the wrong bucket.
The fix is not a better parser. It is a review step that reads the source line, which is why tools that show where each figure came from are easier to review than tools that only show the result.
| Line | As parsed | After review |
|---|---|---|
| Trailing NOI | $1,020,000 | $960,000 |
| Implied value at 6.0% | $17,000,000 | $16,000,000 |
| Difference | — | $1,000,000 |
Before you trust the output
Treat AI output as a first draft by a fast junior analyst. Before it drives an offer or an investment committee memo:
- Keep the source documents attached to the model, and make every key figure traceable to a page or a row.
- Tie totals first: units, square feet, contract rent, the T-12 annual total. If the totals do not tie, stop there.
- Read the lines that move value: anything above a threshold you set, every "other income" line and every one-off.
- Replace seller costs with yours, and re-underwrite property taxes and insurance at your basis.
- Know where the comps came from, what date they are, and whether you are allowed to share them with investors.
- Record who reviewed the inputs and when. If an investor later asks how you got to the price, that record is your answer.
- Remember that the model is not the offering. Projections you show investors carry the same disclosure duties as any other statement in the raise.
The tools, as their vendors list them
The table below is what each vendor publishes on its own website, read on 2026-10-04. It is not a ranking and not a test: we did not run the products. Prices change, so check the source before you rely on a number. CREAssets is our product and is labelled that way.
Two names often appear in this category and are not in the table. Walker & Dunlop’s about page states that in 2019 it "acquired Enodo, a fully automated underwriting platform for multifamily real estate"; we found no current page offering Enodo for sale on its own. Northspyre lists acquisition underwriting as a use case, but its site centres on development budgets, draws and schedules, and we found no published OM, rent roll or T-12 parsing.
| Tool | Built for (per the vendor) | What the vendor publishes | Pricing | Source |
|---|---|---|---|---|
| CREAssets (our product) | Commercial real estate sponsors who raise from investors | OM, rent roll and T-12 parsing from PDF and Excel with OCR fallback, into a typed workbook; scenarios, sensitivities and loan sizing; HUD fair-market-rent benchmarks plus AI-suggested rent comps; the workbook converts into the deal | $0 to start, workbook filled by hand; AI parsing on the Platform plan, $299/mo, within a monthly AI budget per firm | creassets.com/pricing, 2026-10-04 |
| Archer | Multifamily investors, brokers, lenders and property managers; "Multifamily only: conventional, student, and affordable" | Rent roll and T12 parsing "in under 10 seconds"; your own Excel model "loaded in about 1 minute"; "100K+ live rent comps and 100K+ live expense comps"; deal pipeline and sourcing | "From $500 per month", custom quote: annual platform fee plus usage, deals or unlimited; free trial (credit card required); free access on a deal shared with you | archer.re and archer.re/pricing, 2026-10-04 |
| Cactus | Multifamily and self-storage acquisitions and new development; all asset classes on its Max tier | Document extraction (OM, rent roll, T12); Excel export "with formulas intact"; each number traced to its source; market intelligence; an AI Underwriting Agent on its Plus tier | Not published (Core, Plus and Max tiers named); 7-day free trial | trycactus.com and trycactus.com/pricing, 2026-10-04 |
| Clik.ai | Lenders and other capital providers: "the institutions that move capital in commercial real estate" | "AI-powered spreading of T12s, rent rolls, and operating statements into production-ready financial models"; Fannie Mae and Freddie Mac workbook population; lease abstraction; states "99% accuracy across all financial documents" | Not published | clik.ai, 2026-10-04 |
| Blooma | CRE lenders, from small teams to large institutions | Origination Intelligence: document parsing, deal scoring and market data integration; Portfolio Intelligence for loan monitoring; Pro and Enterprise plans | Not published | blooma.ai and blooma.ai/plans, 2026-10-04 |
| Dealpath | Institutional deal teams; plans "typically include a minimum of five users" | AI Extract ingests structured data from OMs and flyers; AI Deal Screening builds a tear sheet from OMs, rent rolls, T12s, BOVs and pro formas; AI Comps from your database, MSCI RCA and third parties; AI Excel Assistant; an Underwriting Scenarios agent listed as coming | Not published | dealpath.com/ai and dealpath.com/plans, 2026-10-04 |
| Cash Flow Portal | Syndicators and fund managers (investor management platform) | AI parsing that turns rent rolls, T-12s and OMs into editable data; scenarios, waterfalls and sensitivity analysis; 2 AI Underwriting seats on its Syndication Suite | Syndication Suite from $699/mo billed yearly | cashflowportal.com/pricing, 2026-10-04 |
Where each one is strong
Each of these products is ahead of ours somewhere. From their own pages:
- Archer publishes 100K+ multifamily rent and expense comps built from parsed financials and lets you keep underwriting in your own Excel model. If your team will not give up its model, that matters.
- Cactus traces every extracted number to its page in the source document, exports Excel with working formulas, and lists comps, tax reassessments and insurance cost estimates in its market intelligence.
- Clik.ai and Blooma are built for lenders, with agency workbooks, loan onboarding and portfolio monitoring that an equity sponsor’s tool does not try to cover.
- Dealpath is a deal-management platform for institutional teams, lists SOC 2 Type 2 security and comps from MSCI RCA alongside your own database, and connects to Excel and outside AI tools.
- Cash Flow Portal puts AI underwriting inside an investor-management platform, as we do, publishes its price, and lets you share underwriting with brokers, lenders and LPs.
- CREAssets has no proprietary comps database beyond public HUD benchmarks and AI-suggested comps, fills its own workbook rather than your Excel model, and holds no security certification.
How to choose
Run a trial on three of your own past deals, including one with a messy scanned T-12, and ask:
- Who is it for? Lender tools, institutional deal-management tools and sponsor platforms solve different problems.
- Which asset types does the vendor list? Several tools here are multifamily-first.
- Can you see the source of each figure, and edit it, before it reaches the model?
- Does it fill your model, its model, or both? What does the Excel export contain: values or formulas?
- Where do the comps come from, how fresh are they, and can you show them to investors?
- What happens after underwriting? If you syndicate, check whether the approved model becomes the deal record or gets retyped into another system.
- What does it cost at your volume? Several vendors price on deals, users or portfolio size and do not publish amounts.
- Who owns the extracted data, and can you export all of it if you leave?