Underwrote.AI is built for hospitality real estate underwriting. Below is an honest accounting of what the engine models, where it’s opinionated or approximate, and what we’re working on next. We update this page when material capabilities ship.
If you’re evaluating Underwrote.AI for a deal that needs something we don’t yet model, email support@underwrote.ai. Customer signal drives roadmap priority.
Hospitality only.Hotels — select-service, full-service, extended-stay, lifestyle, and luxury. We do not currently model multifamily, office, retail, industrial, or specialty CRE. We chose to lead with hospitality because the underwriting conventions (USALI, STR, Smith Travel benchmarks, sub-brand franchise fees) are well-defined enough that a single tool can do them right. Multi-asset support is a question we’ll revisit when hospitality users start asking for it.
The proforma engine is a hybrid of an Excel template (parsed and evaluated per-request via hyperformula) and a JS calculation layer that fills gaps and overlays sponsor-specific overrides. This trusted-math story matters: the same formulas an analyst would write in Excel run on the server, and the Excel export gives you live-formula round-trip.
One template today. The engine ships with a single underwriting framework derived from a real institutional deal (Alley North Hotel). It generalizes well across hotel deals — select-service ground-up Sun Belt, full-service acquisition, adaptive reuse, repositioning — but power users with strongly opinionated workflows may find the wizard inputs over- or under-specified for their style. Multi-template support is on the roadmap.
10-year proforma with departmental P&L per USALI categories (rooms, F&B, other operated departments, undistributed, fixed charges). Five waterfall structures (Simple Split / Pref + Split / Pref + ROC + Split / IRR Lookback / Catch-Up) with engine-vs-Excel parity within $1/year on per-bucket distributions. Three-bucket waterfalls (LP + OZ + GP) supported for opportunity zone deals.
Senior debt with optional refinance event mid-hold. Mezzanine debt with closed-form annuity P&I and exit fee. Preferred equity with current / accrued / PIK behaviors. OZ equity with per-rank pref allocation (pari-passu / OZ-senior / LP-senior). Sponsor co-invest with promote.
Drag-and-drop extraction for STR Reports, Kalibri Reports, Operating P&L PDFs, and existing proforma xlsx files. The AI reads the document and pre-fills the wizard with structured data + per-field confidence scores; you review and accept or edit before anything saves.
Live-formula Excel export (8 visible sheets including live Waterfall, Sensitivity, Scenarios cascade). AI-generated IC memos with PDF export, per-section editing, and stale-memo detection. Returns headline metrics with full decomposition (cashflow stream, IRR, equity multiple, DSCR, Sources & Uses). Workspace branding (logo + company name + IC memo signature) flows into both the IC memo PDF and the Excel cover sheet on Pro plans.
Beyond rooms, F&B, events, and parking, the engine models five additional USALI Other Operated Departments: Spa & Wellness (treatment-room driven with retail attach),Retail (% of rooms revenue for hotel sundries and gift shops), Golf (rounds × $/round + pro shop attach, gated by a hasGolfCourse toggle),Telecom (per-occupied-room-night, often a net cost in modern hotels), and Internet(per-occupied-room-night, gated by isAdditionalCharge — most modern hotels bundle internet into rooms revenue). All five default to tier-appropriate values that you can override.
For luxury and resort deals with private membership clubs (Soho House, Aman Club, Six Senses Place, NoMad-style social clubs), the engine models the four-component revenue mix: amortized initiation fees, escalating annual dues, mandatory F&B minimums (with industry-standard 85% capture rate), and guest visit fees. Capital reserve contributions flow as a separate post-NOI line, not as an operating expense. Member ramp can be flat-curve or cohort-based (with retention).
We do not model club-specific marketing spend, member churn forecasts, or club-driven brand value. We model the steady-state economics; you provide the member-count assumptions.
For mixed-use deals with branded condo-residences, the engine models cost basis allocation between hotel and residential components, sell-out revenue recognition with cost-of-goods-sold recovery, optional rental program economics (residences pooled into hotel rooms inventory), and HOA reserves. Sell-out velocity has three preset curves: aggressive (front-loaded), baseline (ramped), conservative (extended). Debt sizing applies to the hotel cost basis only by default; sponsors can override to 'blended' for deals where senior debt covers both components.
Sell-out velocity is the most uncertain assumption in your deal. The presets reflect typical patterns; actual sell-outs vary widely by market, brand strength, and pre-launch reservations. Underwrote.AI does notpredict your actual sales velocity — it computes the deal economics that follow from your assumption. Sponsors who treat the model output as a sales forecast are misreading the engine.
We do notmodel individual unit-level pricing variation, owner-occupied vs rental ratios over time, or HOA reserve adequacy projections. Standalone for-sale residential development without a hotel component is permanently out of scope — that’s a multifamily- development product, not hospitality underwriting.
Settings is the sixth nav item, accessible from any screen. Six sub-tabs:
Pricing: Lite (free, unlimited deals) · Pro Monthly $249/mo or Annual $1,990/yr (save $998/yr, ~33%). Single Export $79 unlocks the live Waterfall sheet for one specific deal — no subscription, snapshot at purchase.
Underwrote.AI is built to stay valuable when you’re also using Claude or other horizontal AI tools. The deal record stays the source of truth no matter where the analysis happens.
The MCP (Model Context Protocol) server lets Claude — Desktop, Code, or any MCP-compatible client — read your workspace’s deals, comparables, underwriting defaults, and audit history. Token-based authentication; per-token rate limit; revoke at any time from Settings → Integrations. Read-only in this release; write-back ships in a future release. No data leaves your workspace except in response to your authenticated requests.
Capture what actually happened to a deal — closed (price, dev cost, T12 NOI) or killed (reason, stage, competing winner). Workspace-private calibration metrics surface on the Proforma Calibration tab once you’ve captured 3+ outcomes. Cross- workspace aggregate calibration is intentionally not shipped — it requires legal review + explicit opt-in and lands in a future release.
Brokers, lenders, and advisors can sign up under broker mode (free) and hand off listings to sponsors. The sponsor accepts via email link; the deal lands in their workspace with the broker recorded as referrer. Verified Partner status (5+ accepted handoffs + 1+ closed deal) signals trust on the recipient sponsor’s end.
F&B operating cost is computed as F&B revenue × food cost % × 3. The 3× multiplier is an approximation for Cost-of-Goods-Sold + labor + other operating expense (well- worn industry shorthand). For deals with material F&B revenue or atypical F&B economics, override the line directly or treat the result as directional. A more granular F&B department schedule is a future engine refactor.
Both lines escalate at a fixed annual rate (3%/year by default, user-overridable). The engine does not model jurisdiction- specific reassessment rules. In practice this matters for:
Override the property tax line per deal to reflect your jurisdiction’s reality. Jurisdiction-aware tax modeling is on the roadmap when customer demand justifies the complexity.
The engine models one refinance event (e.g., a construction- takeout to permanent at year 2, or a cash-out refi at year 6). Deals that refi twice — common for sub-stabilization construction takeouts followed by mid-hold cash-out — can’t be modeled directly today. The architecture supports generalizing to N events; we ship it when a real deal needs it.
The engine fully models event/banquet revenue and parking revenue, including departmental cost margins by hotel tier. It does not currently model: telecom, spa, golf, retail, internet, or other USALI OOD lines. For deals where these materially contribute to NOI (a resort with a golf course or spa, for example), the impact must be modeled via the wizard’s “Other Revenue” field with a custom margin override. Adding first-class OOD lines is straightforward when buyer signal surfaces.
The engine does not yet thread:
These are real and consequential for adaptive-reuse and opportunity-zone deals. We’re holding them for Phase O+ engineering when we have a real adaptive-reuse customer to calibrate against.
If a deal originates with no senior debt (all-equity) and adds debt at a refi event, the engine synthesizes a phantom origination loan at 65% LTV for the cash-out math. The wizard surfaces this with an italic note in the refinance event card (“no senior debt at origination — falling back to default 65% LTV for cash-out math”). Pre-refi years carry phantom debt service; the user-visible note is sufficient disclosure for now. Fully resolving this requires either an explicit “all equity” affordance or accepting a zero-amount loan in the resolver.
Today the waterfall collapses to pari passu when the GP has 0% equity in the deal. Real-world structures where a sponsor with $0 co-invest still earns promote — common in emerging-GP sponsorship — aren’t supported. We can add a gpEarnsPromoteWithoutCoinvest flag if a real deal needs it.
When sale proceeds are insufficient to satisfy return-of- capital + accrued pref (a deal with severe asset depreciation), the engine pays both in full and runs the residual negative. This produces a small over-distribution at exit visible to anyone who ties LP+GP back to source dollars. We documented this as a known modeling decision rather than fix it speculatively — the right cap (proportional? LP-first? per-structure variant?) depends on what real deals need.
For ground-up deals, the wizard exposes a landCostfield, but the engine’s template-driven Budget tab currently ignores it (we zero acquisitionCostfor groundUp and the template uses that as the source-of-truth land line). Ground-up deals with material land cost should override the Sources & Uses Land line directly. We’ll wire landCost through in a near-term engine pass.
The IRR sensitivity matrix clamps exit cap rate at a 0.5% floor to avoid degenerate terminal values from sub-1% caps. If you’re underwriting trophy assets at sub-1% exit caps, the sensitivity tab won’t reflect cap rates below 0.5%. Underwrite the asset directly (the proforma itself accepts any cap rate); the matrix is a stress-testing surface.
Underwrote.AI is an underwriting tool, not an asset management one. The engine produces a 10-year forward forecast; it does not currently compare actuals against the underwriting baseline. Variance tracking, quarterly reforecast, and budget-vs-actual reporting are deliberately out of scope until we’ve nailed the underwriting flow.
We don’t pre-announce roadmap publicly. Capabilities ship when they’re ready and we update this page. Buyers with specific needs are the strongest signal we can act on — email support@underwrote.ai and tell us what would unlock your deal flow.
A few directional notes for context:
A few conventions that affect how outputs read:
If a number looks wrong, an export looks broken, or you hit a deal shape we don’t handle, email support@underwrote.ai with the deal ID (visible in the URL) and a brief description. We respond within one business day.