Is AI Pricing Legal After the RealPage Settlement?

"My board asked if AI is even legal now"

A manager forwarded me that exact line last month. Her board read a headline about the RealPage settlement and asked whether the company should pull every AI tool it uses, including the one that answers after-hours resident calls. That is the overcorrection happening across the industry right now.

The fear is understandable and the response is wrong. Operators are treating "AI" as one undifferentiated legal risk when the settlement drew a very specific line around one very specific behavior. Freezing your whole stack because of it is like banning all spreadsheets because someone once used one to fix prices.

Here is the uncomfortable part: most of the AI a property management company actually needs has nothing to do with rent-setting at all. The busywork that eats your staff (work orders, records requests, COI tracking, board minutes) is nowhere near the conduct the DOJ cared about.

What the RealPage settlement actually prohibits

Antitrust law does not care whether coordination happens over a phone call, a spreadsheet, or a machine learning model. It cares that competitors who should be setting prices independently instead used a shared mechanism fed by each other's confidential numbers.

That is the hinge. Two ingredients had to be present: non-public competitor data flowing into a common system, and pricing recommendations flowing back out to those same competitors. Remove either ingredient and the antitrust theory falls apart.

Notice what is not on the list. Using your own historical data to price your own units is ordinary business. So is answering a resident, triaging a leak, or drafting board minutes. None of those involve competitor data or coordination, which is why lumping them into "RealPage risk" is a category error.

The bright line: pricing coordination vs. operations AI

There is a clean way to sort any AI tool in your stack. Ask two questions: Does it ingest non-public data from competing landlords? Does it push pricing decisions back to those competitors? If both answers are yes, you are in the danger zone. If either is no, you are almost certainly not.

This is the contrarian point most nervous operators miss: the settlement is narrow, and the AI that actually moves your margin lives entirely outside it. Agents like Mason Maintenance for work order triage or Victor Vendors for COI tracking never touch a competitor's numbers, because there are no competitor numbers involved in dispatching a plumber.

Definition: Operations AI is software that automates documented, deadline-driven administrative work using an organization's own internal data, with humans retaining approval over decisions. It sets no market prices and shares no data with competitors, which places it outside the RealPage theory of harm entirely.

Red-flag vs. safe: audit your AI stack

Run every AI tool you use through this checklist. The red-flag items are worth a lawyer's eyes. The safe items are the ones your board is needlessly panicking about.

Why own-data, single-tenant AI sidesteps the whole risk

"Trained on your own communities, walled from competitors' data" is not marketing language. It is the exact structural fact that keeps an operations agent outside the antitrust question, because the entire RealPage theory requires a shared pool of competitor data that a single-tenant, own-data system never creates.

Single-tenant means the agent learns from your communities, your work order history, your governing documents, your vendor roster, and nothing else. It cannot recommend a rent based on what a competing landlord across town is charging, because it has never seen that number and has no path to it.

This is a design decision with legal consequences. When One Home Agent builds an operations agent for a management company, the training data is the company's own institutional memory. There is no cross-company data lake feeding pricing back to competitors, which is the specific machinery the settlement dismantled.

“The question boards should ask is not "is this AI?" It is "where does its data come from and what does it decide?" An agent trained only on your own communities that never sets a market price is not the thing the DOJ went after. Confusing the two is costing operators the easy wins.”

Todd Paton, Partner, One Home Agent

The disclosure and documentation you should keep anyway

Even for clearly safe operations AI, write down what it does. Not because it is legally radioactive, but because a nervous board, a picky owner, or a future auditor will ask, and a one-page answer beats a scramble.

Good documentation names each tool, states its data source in one line, states what it decides versus what a human approves, and confirms it does not share data with competitors. That single artifact converts "is this even legal?" into a five-minute conversation.

1. 01

   Inventory every AI tool

   List each AI or automation in your stack, including anything staff quietly use on their own. You cannot document what you have not admitted exists.

2. 02

   Record the data source per tool

   For each one, write a single sentence: what data trains it or feeds it. Flag anything that touches competitor data for legal review.

3. 03

   Record the decision boundary

   Note what the tool decides automatically and what requires human sign-off. Consequential actions should sit behind an approval gate.

4. 04

   Confirm the data wall

   For operations agents, confirm in writing that the tool runs on your data only and shares nothing with competing operators. Get it from the vendor if you did not build it.

5. 05

   Keep a disclosure on file

   Draft a short board-facing and resident-facing note describing how you use AI. Having it ready builds trust and defuses the RealPage panic before it starts.

Where operators go from here

The practical move is to separate pricing questions (worth caution and counsel) from operations questions (worth adopting fast). The operations side is where the labor savings live and where the antitrust exposure does not.