Underwrite turns broker material into a deterministic CRE model, a reviewable IC memo, and 138 live Phase 3 modules you can trace back to source code. Fast is useful. Auditable is the product.
Analysts are not short on spreadsheets. They are short on time, audit trails, and clean ways to defend assumptions when an IC asks where a number came from.
A CRE underwriting workbench built around three pillars: source-aware extraction, deterministic analytics, and a memo that reads like an institutional deliverable.
138 live Phase 3 modules cover macro context, sector demand, risk, capital stack, operations, sponsor economics, treasury spread, and rent-roll diagnostics. The math is regression-locked and reviewable.
Paste text or upload deal material. Underwrite extracts assumptions, tenant rent rolls, and source quotes, then hands the numbers to the engine rather than inventing the model.
The output is not a toy summary. It includes Monte Carlo, Deal Strength, Capital Stack visualization, lender risk, stress tests, GP Promote Yield, and recommendation logic.
Built for the moment between broker email and investment committee.
Use the Cherry Street demo, open an empty model, or paste broker materials into the AI parser. Every field remains editable.
Underwrite calculates returns, debt sizing, tax overlays, refinance events, tenant concentration, downside risk, and property-type-aware validation flags.
Open the IC memo, collapse to Executive view, print to PDF, or export Excel. The goal is a package you can inspect, challenge, and defend.
The memo is the proof. It surfaces the analysis by name, so a reviewer can jump from executive summary to risk, capital structure, operations, rent roll, and diagnostics without hunting through a spreadsheet.
Composite 0-100 signal across returns, risk, leverage, and operating quality.
1,000-trial view of return dispersion, downside probability, and tail stats.
Year-by-year covenant cushion at base, -10%, -15%, and -20% NOI.
Single-asset volatility compared to a 25-deal diversified portfolio.
Tracks lender exposure, asset coverage, and cash-out refi capacity by year.
Tests lender sizing and cash flow durability under higher refinance rates.
Flags approval posture, parking relief, adaptive reuse, density sensitivity, and counsel-review triggers.
Triages prior-use history, Phase II probability, vapor risk, floodplain exposure, and diligence delay.
Converts site-condition risk into reserve need, capacity, funding gap, delay, and lender escrow read.
Shows senior debt, mezzanine debt, sponsor equity, and LP equity in one view.
Compares project IRR to blended capital cost across debt and equity.
Separates operating yield, debt paydown, and appreciation into a stacked view.
Inflation-adjusted IRR and cap-rate framing for purchasing-power review.
Shows how much NOI converts into distributable cash after debt and capex.
Measures how revenue misses translate into NOI misses by cost structure.
Compares insurance as bps of value and per-unit cost to asset-class bands.
Classifies carrier availability, renewal shock, deductible reset, and lender exception risk.
Turns peril severity into deductible dollars, NOI drag, reserve coverage, and exclusion sensitivity.
Synthesizes placement route, escrow posture, broker timing, and lender conversation points.
Tests whether insurance growth is tenant-recoverable or sponsor-absorbed NOI drag.
Screens tax, insurance, CAM, utilities, and repair pass-through language for opex leakage.
Frames reassessment shock, appeal evidence, reserve coverage, and uncovered tax exposure.
Tests electric, gas, water, sewer, and common-area utility growth for recoverability and NOI drag.
Classifies expense variance as controllable, recoverable, reserve-funded, appealable, or sponsor-absorbed.
Tests recurring maintenance, reserves, asset age, inflation, and deferred-maintenance funding gaps.
Shows stabilized cap-rate movement as NOI grows through the hold.
Ranks largest tenant exposure, top-three share, and rollover concentration.
Separates GP economics from LP return quality and promote compensation.
Scores whether the deal profile fits income, growth, or risk-sensitive LPs.
Frames whether exit or refinance is more defensible under modeled terms.
The point is not to replace judgment. The point is to make the model behind that judgment faster, clearer, and more reviewable.
| Excel models | Legacy templates | Generic AI tools | Underwrite | |
|---|---|---|---|---|
| Model logic | Flexible, but varies by workbook | Static and hard to adapt | Often opaque | Deterministic engine with reviewable methodology |
| Audit trail | Depends on model discipline | Formula tabs, limited narrative | Weak unless manually documented | Source quotes, validation flags, memo sections |
| Depth | Whatever the analyst builds | Template-specific | Broad but shallow | 138 live Phase 3 modules across macro, risk, debt, ops, LP |
| Confidentiality posture | Local files, inconsistent sharing | Depends on template workflow | Easy to overshare sensitive data | Confidential Mode and local-first deal controls |
| Output | Workbook first | Workbook plus manual memo work | Narrative without defensible math | Excel export, IC memo, portfolio comparison |
Four ways to start. New public access is paid from the first active account.
I built Underwrite because CRE underwriting still assumes a fresh blank workbook every Monday. That is not rigor. That is typing.
The standard should be higher: deterministic math, explicit assumptions, source-aware extraction, and a memo that can survive review from someone who knows the business.
Underwrite is my attempt to put a stake in the ground. AI can help analysts move faster, but the model still has to be inspectable. The numbers have to tie. The reviewer has to be able to ask why.
If you have ever lost a weekend to a deal that did not close, you already know why this exists.
New accounts start with a paid beta plan. Prior written Founding 100 commitments remain honored. No automatic charge. Ali replies personally.