Real Estate & Property

The deal lives in a spreadsheet. So does the mistake

Commercial real estate runs on Excel and has resisted every attempt to move it elsewhere, because every deal is bespoke. SheetAI works inside the model you already built: cleaning rent rolls, abstracting leases into rows, checking that your NOI actually reconciles, and writes standard formulas an investment committee can audit line by line.

Try it on your sheet See real workflows

Audit-ready trailNo data trainingReversible per action

91%

Of audited spreadsheets had errors

Across the most rigorous field audits of working spreadsheets, 91% of the 55 examined contained serious errors. Compiled by Raymond Panko, University of Hawaii, the standard reference on spreadsheet error research.

Source: Raymond Panko: Audits of Operational Spreadsheets

Have met all their AI goals

88% of real estate investors, owners and landlords have started piloting AI, but only 5% say they have achieved all their programme goals. JLL surveyed more than 1,500 senior decision-makers in 16 markets.

Source: JLL: Global Real Estate Technology Survey (2025)

2 in 3

Rank data fluency a key leader skill

Two in three commercial real estate executives say AI and data fluency will be among the most important capabilities for future leaders. Deloitte surveyed 950 C-level executives in mid-2026.

Source: Deloitte: 2027 Commercial Real Estate Outlook

Real Estate & Property: illustration for SheetAI's real estate & property use case

Commercial real estate still underwrites in Excel, and the industry knows the risk.

Altus Group's CRE innovation research has consistently found spreadsheets remain the primary tool for valuation and underwriting, with data fragmentation cited as the sector's leading operational challenge. The models are bespoke, the inputs arrive as PDFs, and the checking is manual.

Source: Altus Group: CRE Innovation Research

This page is about the analytical work behind property transactions and management: comparable sales and rent analysis, rent roll normalisation, lease abstraction, acquisition underwriting, debt sizing, portfolio reporting, and the monthly owner statements that nobody enjoys producing. SheetAI reads the ranges in your model, writes standard Excel and Google Sheets formulas, and produces the checks that separate a defensible underwrite from an optimistic one. It does not value property, it does not give investment advice, and it is not a substitute for a qualified appraisal; the boundaries section is exact about that, because in this sector the difference between analysis and valuation is a regulated one.

The state of property analysis in 2026

Four realities that explain why the sector's most important decisions still get made in a workbook.

CRE firms citing spreadsheets as their primary valuation tool

Majority

Purpose-built platforms have made real inroads in portfolio management and reporting, but deal-level underwriting stays in Excel because every deal has a structure the platform did not anticipate.

Altus Group CRE surveys

Data fragmentation cited as a leading operational challenge

Top-ranked

Rent rolls from the property manager, operating statements from the accountant, leases as PDFs, comps from a subscription service: four sources, four formats, one model.

Altus Group / industry surveys

Spreadsheets containing at least one error

~90%

Audit studies across sectors converge on this figure. In underwriting, the error that matters is usually an input carried forward from the last deal rather than a broken formula.

EuSpRIG

Residential transactions where pricing is set from comparables

Nearly all

Comparable analysis is the foundation of residential valuation, and the selection of which comparables to use is where judgement (and bias) enters.

Standard appraisal practice

The pattern: the sector's core analytical work is document-to-model transcription and structured comparison: reading a lease into rows, reconciling a rent roll against an operating statement, checking a comp set for like-for-like. That is precisely the work an AI inside the spreadsheet can take on without asking anyone to abandon the model they trust.

Anatomy of an acquisition underwrite

Where the week actually goes on a mid-market commercial acquisition, before any of it is automated.

Phase 1: Assemble the data room

›Rent roll from the property manager
›Trailing 12 operating statements
›Lease PDFs, service charge budgets, capex history

⚠ Four formats, none of which agree on tenant naming, unit references, or period boundaries. Reconciling them is the first two days.

Phase 2: Normalise the rent roll

›Standardise unit references and tenant names
›Convert rents to a common basis and period
›Flag vacancies, holdovers, and expiring terms

⚠ The rent roll as supplied never ties to the operating statement, and finding out why is a line-by-line exercise nobody budgets for.

Phase 3: Abstract the leases

›Term, commencement, expiry, break options
›Rent steps, reviews, and indexation
›Recovery structure, caps, and exclusions

⚠ An hour a lease, done under time pressure, and the option nobody spotted is the one that resets the exit assumption.

Phase 4: Build and stress the model

›NOI build-up from the normalised roll
›Debt sizing against DSCR and LTV covenants
›Sensitivity on exit yield, rent growth, and vacancy

⚠ The model is last deal's model with the numbers changed, which means it inherits last deal's assumptions and last deal's errors.

Phase 5: Package for committee

›Summary metrics, comp set, and risk register
›Reconcile every headline number to source
›Answer the question about cell F212

⚠ Provenance is what committee actually tests, and it is the thing least likely to have been recorded while the model was being built at speed.

Cautionary tale

One input, carried forward

In 2012 the London Whale losses at JPMorgan ran past $6 billion, and the bank's own task force report traced part of the failure to a value-at-risk model implemented across spreadsheets where an analyst copied and pasted data between worksheets by hand. One operational error (dividing by a sum where an average was intended) understated the modelled risk by roughly half. The individual mistake was trivial. What made it consequential was that the model had been rolled forward, trusted, and never independently re-derived.

The lesson: Every underwriting model in this sector is a copy of a previous one, which means every assumption in it has a provenance nobody has checked since the deal it came from. The practical defence is to make the model state its own inputs: where each number came from, which are hardcoded, which are linked, and which changed since the last version. SheetAI will produce exactly that audit: listing hardcoded values inside formula ranges, flagging inputs that differ from the source documents, and reconciling the headline metrics back to the rent roll.

Source: JPMorgan Chase: Report of the Management Task Force Regarding 2012 CIO Losses

Who it's for

If your deal, your portfolio, or your month-end lives in a workbook, this is the page.

CRE Acquisition Analysts

Rebuild an underwriting model per deal under a deadline, inheriting the last one's assumptions.

Property Managers

Produce owner statements and rent rolls monthly across a portfolio of assets and systems.

Residential Agents & Teams

Assemble comparable analyses per instruction, by hand, from a subscription export.

Asset Managers

Track lease events, break options, and expiries across hundreds of tenancies in a spreadsheet.

Investors & Syndicators

Need a defensible model and an investor pack that ties back to source documents.

Real property workflows

The prompt, what the agent does with your model, and the state it leaves it in.

real-estate_workbook.xlsx - SheetAI

You ask

“RentRoll!A:M came from the property manager and T12!A:H is the trailing operating statement. Standardise unit references and tenant names in the rent roll, annualise the rents onto a common basis, then reconcile total contractual rent against the T12 rental income line. List every difference over £2,000 with a probable cause.”

SheetAI does

✓Normalises unit and tenant identifiers so the two sources join.
✓Converts rents onto a common annual basis regardless of how each line was stated.
✓Reconciles the totals and computes line-level variances.
✓Characterises each material difference: vacancy timing, rent free, holdover, arrears.

Formula written

=SUMIFS(RentRoll!G:G,RentRoll!C:C,A2)-SUMIFS(T12!D:D,T12!B:B,A2)

Result

Rent roll normalised across 84 units and reconciled to within £900 of the T12. Six differences explained, two flagged as unexplained, both of which turned out to be side letters absent from the data room.

Everything property teams need, inside the model they already use

Rent roll normalisation, lease abstraction with provenance, comp screening with documented exclusions, and a model audit before anything reaches committee.

✓Rent roll and operating statement reconciliation

✓Lease abstraction with page-level sources

✓Comparable screening with documented exclusions

✓NOI build-up and DSCR / LTV debt sizing

✓Two-variable sensitivity grids

✓Workbook audit for hardcodes and broken patterns

Plays well with your stack

Yardi and MRI exports
AppFolio and Buildium reports
Argus cash-flow exports
MLS and portal comparable downloads
Lease PDFs and data-room documents
Accounting exports (Xero, QuickBooks, Sage)

What one acquisition underwrite looks like

A representative mid-market commercial acquisition (84 units, 14 leases in the data room) before and after the transcription and checking move into the sheet.

Before SheetAI

~52 hours

Mon
Assemble the data room; reconcile four incompatible formats
~9h
Tue
Normalise the rent roll; chase why it does not tie to the T12
~11h
Wed
Abstract 14 leases by hand into the tracker
~12h
Thu
Build the NOI, size the debt, run sensitivities manually
~11h
Fri
Package for committee; re-derive the numbers under challenge
~9h

✓

With SheetAI

~14 hours

Mon
Rent roll normalised and reconciled; variances explained
~3h
Tue
Leases abstracted with provenance; REVIEW items to counsel
~3h
Wed
NOI and debt sizing built; sensitivity grid generated
~3h
Thu
Model audited; 31 findings cleared before committee
~3h
Fri
Pack assembled with every number traced to source
~2h

~73% reduction on a representative 84-unit acquisition underwrite.

What SheetAI will not do

Property is a sector where analysis and valuation are legally distinct activities. The line is worth stating precisely.

Value a property

Producing a formal valuation is a regulated professional activity under RICS Red Book standards in the UK, USPAP in the US, and comparable regimes elsewhere. SheetAI computes comparables, ratios, and cash flows from inputs you supply. It does not opine on market value, and no output from it should be represented as a valuation to a lender, a court, or an investor.

Give investment advice

It will not tell you whether to buy, at what price, or on what terms. It builds and checks the analysis that informs that judgement. Where a recommendation constitutes regulated financial advice, it belongs with someone authorised to give it.

Interpret a lease as a lawyer would

It abstracts commercial terms into rows and flags ambiguity rather than resolving it. Whether a break notice was validly served, whether a review mechanism is upward-only in effect as well as in drafting, and what a poorly drafted recovery clause actually means are legal questions. The REVIEW flags exist precisely so those reach a solicitor.

Replace due diligence

It reconciles what is in the data room. It cannot tell you what is missing from it: the side letter that was never uploaded, the dispute nobody disclosed, the structural survey that has not been commissioned. Reconciliation is not diligence, and the gap between them is where deals go wrong.

“

The lease abstraction used to be a full day per deal and it was the part I trusted least, because you are reading fourteen documents at nine at night. Now the terms land in the tracker with a page reference and the ambiguous ones are flagged. We found a break option in the second deal we ran it on that had been missed in the original underwrite.

Investment Analyst, mid-market CRE fund (£180m AUM)

Frequently asked

What property professionals ask before they let a model near a live deal.

Can it value a property?

›

No, and this matters more in real estate than almost anywhere else. Formal valuation is a regulated professional activity (RICS Red Book in the UK, USPAP in the US) and requires a qualified valuer's opinion and signature. SheetAI computes comparables, per-unit metrics, yields, and cash flows from the inputs you give it. That is analysis, and it is genuinely useful, but it is not a valuation and must never be presented to a lender or a court as one.

Will it work with my existing underwriting model?

›

Yes. That is the point. It reads the ranges in the workbook you already built rather than asking you to adopt a template. Every deal in this sector has a structure the platform did not anticipate, which is why the model has survived every attempt to replace it. The formulas it writes are standard Excel and Google Sheets syntax, so the model stays portable and auditable.

How well does lease abstraction work?

›

For commercially drafted leases with a text layer, extraction of the core terms (parties, dates, rent, review mechanism, recovery basis) is reliable, and every field comes with the source page so it can be checked in seconds. What it deliberately does not do is resolve ambiguous drafting: those terms come back marked REVIEW. Given that a missed break option can reset an exit assumption entirely, a flagged uncertainty is worth considerably more than a confident guess.

Can it pull comparables for me?

›

It can research publicly published listings and sold-price data and screen a comp set you supply against explicit criteria. What it cannot do is access your MLS or subscription comparables service on your behalf; those carry licence terms about redistribution and automated access that are between you and the provider. Export from your service, and the screening, exclusion logic, and per-unit analysis happen in the sheet.

Does it handle service charges and recoveries?

›

Yes, and this is where a lot of underwriting error lives. It can build the recovery calculation from the abstracted lease terms, apply caps and exclusions, and reconcile budgeted against recovered amounts. Where the lease drafting on a recovery is ambiguous it flags it, which is the honest answer, because the same clause can be modelled two defensible ways.

Is my deal data confidential?

›

Your workbook stays in your account and nothing is published anywhere. That said, treat data-room material according to whatever confidentiality undertaking you signed; the obligations attached to a data room do not change because the file is open in a different tool. If your NDA restricts processing by third-party services, check it before uploading.

Can it audit a model I inherited?

›

This is one of the highest-value things to ask it. Hardcoded numbers buried inside formula ranges, a formula pattern that breaks in column AK, an external link to a file nobody has access to, an assumption carried forward from a deal two years ago; it will list all of it with cell references and severity. On an inherited model that is an afternoon's work you would otherwise never get budget for.

Does it work for residential as well as commercial?

›

Yes. Commercial work skews toward rent rolls, lease abstraction, and debt sizing; residential toward comparable analysis, portfolio tracking, and refurbishment appraisals. The underlying capability (reading documents into structured rows, screening against explicit criteria, checking the model) applies to both.

Sources & further reading

Every benchmark and statistic on this page is drawn from publicly available research. We cite our sources because we read theirs.

Get started

Stop rebuilding the model.
Start checking it.

Open your underwriting workbook and ask for an audit. You will have the hardcodes, the broken patterns, and the carried-forward assumptions listed before your coffee goes cold.

All use cases

Pay as you go, no subscription required