Why we built a hospitality underwriting tool when the market already has Excel templates and Argus.
Hospitality underwriting has a problem nobody wants to admit: the spreadsheet is the bottleneck.
Every sponsor, family office, and acquirer in the asset class runs the same workflow. An associate inherits a 14-tab Excel model from the deal before. They open the OM, copy numbers, fight broken formulas, paste in a Smith Travel report, recompute the waterfall by hand because a cell reference moved, and ship a workbook to IC that nobody really trusts because nobody can audit every cell in the two hours before the meeting.
We built Underwrote.AI because the math should be a versioned engine, not a copy-pasted artifact. Same USALI categories. Same Smith Travel benchmarks. Same franchise fee tables. Live formulas in the Excel export when you hand it to the lender. Zero broken references when the next deal lands.
Three things, specifically.
A real engine, not a template.The proforma runs on a deterministic calculation layer that powers a live-formula Excel export. The same math runs on the server and in the workbook your IC opens. No mock data. No silent fallback when something fails — errors surface to you, with a retry, by design.
Hospitality-native modeling.Other Operated Departments — spa, golf, retail, telecom, internet, membership clubs, branded residences — are first-class. Five partnership waterfall structures including OZ-equity three-bucket variants. Mezzanine, preferred equity, and refinance with cash-out threaded end-to-end. Built around real institutional deals, not a generic CRE template.
Open to the AI you already use.Underwrote.AI exposes an MCP server that lets Claude, ChatGPT, or Copilot read your workspace data with a revocable token. Your deals stay yours; the AI tools your team already trusts can answer questions grounded in your real models. We don’t lock you into our analytics.
This is the part we want you to read before you sign up.
Underwrote.AI computes the deal economics that follow from your assumptions. It does notpredict outcomes. Sell-out velocity on a branded residence project is the most consequential and most uncertain assumption in the deal — and we say so, in the wizard, in the Underwrote_AI_Brief tab, and on the Scopepage. We give you three preset curves and let you override; we do not tell you which is right for your market. That’s your judgment.
We don’t write IC memos behind your back. The Underwrote_AI_Brief tab hands your data + a hardcoded prompt to Claude or ChatGPT; you watch the narrative generate, you edit it, you ship it. We do not run autonomous agents that produce conclusions you didn’t author.
We don’t model what we haven’t validated. Asset management variance tracking, jurisdiction-specific property tax reassessment, second refinance events, C-PACE / NMTC / HTC threading — all explicitly out of scope until a real deal calibrates them. The full list is on the Scopepage. We’d rather be honest about what we don’t model than ship a feature that produces a confidently wrong number.
A few principles, since pre-launch tools are bets and you should know which bets we’re making.
support@underwrote.ai — product, sales, partnerships, bugs, billing, account help.
security@underwrote.ai — diligence packs, DPAs, vulnerability reports.