Home
News ▼
Business
Entertainment ▼
Lifestyle ▼
Technology ▼
Contact Webmail My Account
Home Editorial & Analysis What Property Investors Can Learn From the Machines Now Underwriting Apartment Blocks

What Property Investors Can Learn From the Machines Now Underwriting Apartment Blocks

What Property Investors Can Learn From the Machines Now Underwriting Apartment Blocks

AI Key Points

  • Traditional multifamily underwriting involves transcribing figures from documents like the offering memorandum, T12, and rent roll into a spreadsheet template, a process that takes days and is prone to errors.
  • Transcription errors such as misclassifying a one-off roof repair as ordinary maintenance or using a mid-month rent roll export can significantly distort a deal's pro forma.
  • Modelling property taxes at the seller's current assessment rather than the post-sale reassessed value is a common reason a pro forma looks better than reality, since the tax line is often among the largest expenses.
  • Because underwriting each deal was labour-intensive, firms could fully model only a small fraction of opportunities, and the good deals they never had time to model represented a hidden cost.
  • Automation is most useful for extracting and normalising inconsistent accounting labels, pulling market comparables into assumptions at the point of use, modelling multiple financing structures in parallel, and maintaining an audit trail of assumptions.

Buying a flat with a mortgage and buying a 240-unit apartment complex with a syndicated loan look like completely different activities. One involves a broker, a valuation and a stack of paperwork. The other involves a twelve-month trailing income statement, a rent roll with hundreds of lease expiries, a debt term sheet and a committee of people who want to know why your exit cap rate is what it is. But underneath both sits the same arithmetic: what does the property earn, what does the debt cost, and what happens if either number moves.

What has changed in the past few years is how quickly that arithmetic gets done on the commercial side. Software that reads deal documents and builds the model for you has moved from novelty to normal practice at a lot of investment firms. The techniques filtering out of that world are worth understanding even if your own portfolio is a buy-to-let or two, because they clarify where property risk actually hides.

The old bottleneck: three days per deal, most of it typing

A traditional multifamily underwriting process starts with a document dump. The broker sends an offering memorandum, a T12 (twelve months of trailing operating statements), a current rent roll, perhaps a tax bill and an insurance declaration page. An analyst opens the spreadsheet template the firm has used for years and begins transcribing: gross potential rent, loss to lease, vacancy, concessions, bad debt, then every expense line from repairs and maintenance to management fees.

Transcription is where the errors live. A T12 that lumps a one-off roof repair into ordinary maintenance inflates the expense base and kills a workable deal. A rent roll exported mid-month can double-count a unit that has turned. Property taxes modelled at the seller's current assessment rather than the reassessed value after sale is the single most common reason a pro forma looks better than reality — in some jurisdictions reassessment on transfer can shift the tax line materially, and that line is often one of the two largest expenses on the statement.

Because each deal consumed so many analyst hours, firms rationed their attention. A team seeing 200 opportunities a quarter might fully underwrite 20 and screen the rest on gut feel and a cap rate scribble. The deals that never got modelled were a silent cost: nobody counts the good acquisition they passed on because they were busy.

What automation actually does well

The useful versions of this technology are not trying to have an opinion about the deal. They are doing four things that humans do slowly and inconsistently.

Extraction and normalisation

Every property manager labels their chart of accounts differently. One calls it "turnover costs", another splits it across painting, cleaning and flooring. Automated extraction maps those variants into a consistent structure, so a portfolio of twelve candidate deals can be compared line against line rather than vibe against vibe.

Market context without a research detour

Rent comps, sale comps and local tax treatment used to mean three browser tabs and a phone call. Pulling that context into the model at the moment the assumption is made changes behaviour: an analyst is far less likely to pencil in 4% rent growth when the comparable set is visible in the same screen.

Financing scenarios in parallel

Debt is where returns are made and lost. Fixed versus floating, five-year versus ten, interest-only period, a rate cap premium, a supplemental loan at year three — each permutation changes the levered return and the downside. Modelling five structures by hand takes a day; modelling them automatically takes the decision out of the realm of guesswork and into comparison.

An audit trail

When an investment committee asks why the insurance line jumped 30%, somebody needs to answer in a sentence, not reverse-engineer a formula written by a colleague who left. Versioned assumptions with visible provenance are unglamorous and enormously valuable.

This is not speculative. Lenders and brokers have been running document-reading underwriting engines on live transactions for several years now; industry coverage of AI underwriting tools used on completed commercial deals showed the approach moving from pilot to production well before the current wave of generative models arrived.

The part most teams still get wrong: closing the loop

Speed is the easy win. The harder discipline is comparing what you assumed at acquisition to what the asset actually delivered. Most firms underwrite a deal, close it, hand the file to asset management and never formally revisit the model. Three years later nobody can say whether the renovation premium was achieved, whether the payroll assumption held, or whether the stabilised vacancy was optimistic by 150 basis points.

Firms that do run that comparison get a compounding advantage, because their assumptions stop being folklore. You learn that your insurance estimates run 10–15% light in coastal submarkets, that your lease-up periods are consistently a month too short, or that your CapEx contingency is about right. That feedback is worth more than any single clever deal.

Where purpose-built platforms fit

Spreadsheets are not going away, and no sensible operator wants a black box deciding what to buy. What the better tools offer is the removal of the clerical layer so the judgement layer gets more oxygen. Platforms such as NOAL approach multifamily real estate underwriting by taking the documents a team already receives — T12s, offering memoranda, rent rolls — and populating a full model from them, with automatic comps, jurisdiction-specific tax context and dedicated expense agents for taxes, insurance and payroll comparisons. Line-by-line CapEx budgeting sits inside the model rather than in a side file; capital stack and equity waterfall modelling runs alongside live loan options and lender comparisons; shared workspaces keep audit trails as assumptions change, and the same model produces IC-ready reporting and proformas before carrying through into post-close asset management. Deals can also be underwritten from an address alone, which matters for off-market sourcing where no document package exists yet.

Borrowing the discipline at a smaller scale

If you own one or two rental properties rather than a fund, three habits from institutional underwriting travel well.

  • Separate recurring expenses from one-off capital spend. A new boiler is not a maintenance cost; mixing them distorts your real yield in both directions.
  • Stress-test the debt, not the rent. Rent is reasonably sticky; refinancing at a materially higher rate is the event that forces sales. Understanding how amortisation, term and rate interact with monthly outgoings is foundational, and the basic mechanics behind how mortgage repayments are actually calculated are the same ones driving a nine-figure apartment loan.
  • Write your assumptions down before you buy, then check them at the twelve-month mark. Void periods, letting fees, insurance renewals, service charges. The exercise takes an hour and corrects a surprising amount of optimism.

The judgement that software cannot supply

A model will tell you what a set of assumptions produces. It will not tell you whether the submarket is about to absorb 1,800 new units, whether the seller's concessions are hiding a management problem, or whether your lender will still be lending in eighteen months. Those remain human calls, informed by walking the property and talking to people who live near it.

What automation changes is the ratio. Less time reconciling a rent roll, more time arguing about exit cap rates and tenant demand — which is the argument that actually determines whether a deal works.