The investment ecosystem
for operating real estate
for operating real estate
One record, many readers. It lives with the property forever.
Phil Bernardo, Founder · Pre-Seed · September 2026
PATL · Confidential
01 / 18
/The Problem
02
The problem
| Who | What they do |
|---|---|
| Broker | Builds the marketing proforma |
| Sponsor | Rebuilds it to bid |
| Lender | Rebuilds it to size debt |
| Insurance | Rebuilds it to quote |
| IC | Rebuilds it to approve |
| Asset management | Rebuilds it to report |
When the deal underperforms, nobody can attribute the miss to the assumption that caused it, because that assumption no longer exists in a form anyone can query.
The model is abandoned at close, the moment it starts to be worth something.
/Why It Is Unsolved
03
Why it is still unsolved
Real estate investment software was built multifamily-first, which means it models leases. Four asset classes do not have them.
| Asset class | Revenue unit | Pricing cadence | What a lease-based engine cannot represent |
|---|---|---|---|
| Hotels | RevPAR per available room | Nightly | USALI departmental P&L, franchise and management fees, PIP capex, brand reserves |
| Short-term rental | Revenue per available night | Nightly, by channel | Channel mix and take rates, cleaning and turnover cost, no tenancy of any kind |
| Self storage | RevPAF per square foot | Monthly, no term | Unit mix, existing customer rate increases, street versus in-place spread |
| Senior housing | RevPAU per unit plus care | Monthly plus acuity | Four licensed products in one building, staffing-driven expense, census by care level |
Everyone builds for multifamily because it is the biggest number on the slide, which is why it is the worst place to enter. An incumbent needs a rewrite to reach these classes. PATL needs a configuration file.
/The Operating Real Estate Ecosystem
04
The operating real estate ecosystem
Documents in, approved assumptions out.
One engine per deal. Every figure traces to an assumption a person approved.
Every run, won or passed, becomes evidence for the next underwriting.
Debt, equity and cover quoted on the numbers the deal was underwritten on.
A standing channel that sees flow before the book goes wide.
The asset held to the plan it was bought on; forecasts re-run off actuals.
Coverage, basis and operating drift read early, as a dated signal.
/Why It Is One System
05
Why it is one system
Signaling finds the asset, brokers bring the book, the Studio prices it, the marketplace funds it, asset management holds the asset to that price, and the data layer makes the next underwriting sharper.
Nothing leaves the wheel. Every turn adds approved assumptions and realised outcomes to the same record, so the next turn costs less than the last — Learn feeds back into Signal.
/Layer 01 · Underwriting Studio
06
Layer 01 · Underwriting Studio
The language model lifts the rent roll, operating statement and STAR data from the OM and tags every value with its source page.
The Python engine builds the USALI departmental proforma, sizes debt on DSCR, debt yield and LTV, and returns a ten-year cash flow.
The coherence engine flags assumptions that disagree with each other or with the market panel. Every override needs a written reason.
The run persists. The broker package, credit memo, IC deck and variance report are views of that run, never copies.
Engine roughly 90% complete, running against live offering memoranda today.
147 brands across 12 franchisors in the fee repository, on 2026 FDD vintages.
Same inputs, same outputs, every run. A test keeps the AI service out of the compute layer.
/The Decision Layer
07
The decision layer
Every proptech deck says AI-powered. PATL moves the language model further from the number.
run_deal
update_assumptions
get_deal_diagnostics
get_brand_fees
get_market_snapshot
list_deals
write_review_memo
MCP tools call the compute layer and never the language model layer. A test enforces it. An agent that asks PATL for a number gets the same number every time, which is what lets parties who do not trust each other share one model.
Agent agnostic
/Layer 04 · Broker Connections
08
Layer 04 · Broker connections
Brokers give early looks to buyers who answer fast and honestly. With extraction and screening already mechanical, a credible answer takes hours.
A proforma builder that works at the listing pitch, before an OM exists.
A live model sent to the buyer list in place of an emailed PDF.
Pass reasons recorded against the run and returned, which earns the next early look.
The broker cannot edit the math. Broker-asserted inputs render apart from engine figures on every export.
Every override carries a written justification that stays on the record.
Neutrality is a property of the engine, enforced by test.
/Layer 03 · Capital & Insurance Marketplace
09
Layer 03 · Capital & insurance marketplace
What changed since version one, who changed it, and why?
The credit committee question · a workbook cannot answer it
/Layer 05 · Asset Management & Forecasting
10
Layer 05 · Asset management & forecasting
Bought at 1.41x DSCR on $91.40 RevPAR and a stabilized $1.84M NOI.
Twenty-six months later, is the asset on its thesis?
/Layers 02 & 06 · The Long Game
11
Layers 02 and 06 · The long game
Every run retained, won or passed, with its approved assumptions attached.
Comparables built from reviewed inputs, queryable by market, brand, keys, vintage and basis.
Improves each time an analyst approves a line, at no added cost.
Coverage, basis, maturity and operating drift watched together across the tracked set.
Thresholds set from underwriting, so a signal means something specific about one asset.
A dated signal trail, so the earliest observation is recoverable when the asset trades.
A license-class flag on every series row separates public federal data from licensed passthrough. This layer is excluded from every projection in this deck.
/Why Now
12
Why now
Reading a 60-page offering memorandum with a departmental operating statement and a STAR report was not reliable two years ago. It is now, which removes the manual step that made this category uneconomic to serve.
Every acquisitions team is being handed an agent and no deterministic tool for operating assets to point it at. That gap is open for exactly as long as nobody builds the engine.
USALI, and its analogs in storage and senior housing, mean the departmental structure is global. The engine does not have to be re-specified per market, only per class.
/Why Me
13
Why me
Fifteen years on three different sides of the same file. Each side taught the product something specific, and each one is a feature you can point at.
Marriott, Universal Orlando, Disney Parks. Ten years pricing the night and running the departmental P&L that the underwriting is supposed to represent.
In the product: the engine models RevPAR the way an operator manages it, not the way a lease-based tool approximates it.
Xenia Hotels and Resorts, Hillpointe, Hyatt. Held assets against assumptions somebody else made and could no longer explain.
In the product: the override justification field. It exists because I needed it and could not have it.
Director of Growth Strategy and Analytics at a major hotel brand today. PivotPt Capital, my own acquisition fund, is the first user of this engine.
In the product: I am the first customer, which is why the commercial gate excludes my own network.
/Business Model
14
Business model
Every tier runs the same engine. What changes is how many assets, how many readers, and how much of the record persists.
Also on the rate card: investor-host seats at $49 a month, the asset management module at $125 per asset a month, and broker listing packages at $2,500 each.
/Distribution
15
Distribution
Pre-revenue. The engine runs against live offering memoranda today, and the first paid accounts are what this round buys.
25 paying accounts, 15 of them outside my network. Network deals prove the product works; the other fifteen prove it sells.
Each layer makes the record more valuable to every party already on it, because the alternative to a shared record is a re-key.
/Market
16
Market
Bottom-up from sourced asset counts, resolved to buying entities rather than buildings: 62,000 US hotels resolve to 17,416 owner entities.
/Risk
17
Risk
| Objection | Answer |
|---|---|
| Six layers is too much surface for a pre-seed | The layers are views over one record that already exists, and the four classes share one engine. The marketplace and signaling layers are unfunded in this round. |
| A foundation model will do this natively | A model can write the arithmetic. It cannot be the system of record, because a record has to return the same answer twice. |
| Argus and Dealpath will move down | Both are lease-based, so reaching these classes is a rewrite. Dealpath shipping MCP validates the category, and that same gap makes incumbents likelier buyers than builders. |
| Brokers will not adopt a neutral tool | Neutrality is what makes the buyer keep the file, and buyers discarding models is the broker's problem. The $99 seat sells against that. |
| Solo founder, no commercial hire | I move full-time at first close. Twelve founder-led months, and the gate is 25 paying accounts, 15 outside my network, before any hire. |
The ask
Post-money SAFE. A $750K minimum close funds twelve founder-led months and the commercial gate without a commercial hire.
The $660K across go-to-market and operations carries both hires, made after the commercial gate clears.
$440K of reach against $100K of code. The engine runs on live offering memoranda today, so this round buys distribution rather than build.