Skip to content
A Realpage framework for AI agents in rental housing and real estate

Open to agents. Governed at the front door.

AI agents are becoming a new kind of worker in rental housing and across real estate. Here is how we welcome them, and how we keep every one of them accountable.

A framework for AI agents in rental housing and real estate

Executive summary

AI agents will increasingly lease, collect, schedule and respond alongside property teams and residents. We welcome them.

That future should be open, with operators free to choose the technologies that work best for their businesses. And it has to be built on trust, because our industry runs on it.

That trust matters more as systems connect. Operators’ platforms are linked to their vendors and to each other, so when one agent misbehaves, the effects don’t stay in one place. They can reach every system, operator and resident connected to it.

But openness requires accountability. Agents must identify themselves, act within authorized permissions and leave a clear record of what they did. That’s the line we draw.

To do their work, agents need to read from and write to the systems operators run on. What we will not accept is unidentified access: agents that sign in with a person’s credentials, operate screens as if they were a person, or disguise who they are. Those agents cannot be identified, limited or stopped, and when something goes wrong, no one can say who is responsible.

This paper sets out seven principles for how agents should be treated and five questions every operator should ask any AI vendor.

Why we’re optimistic

Agents won’t just make this industry faster. They’ll change what it can do.

Software has long helped real estate keep its records. Agents are beginning to do the work itself: leasing, collecting, maintaining, reconciling and analyzing. As they mature, that work will run continuously and at the scale of entire portfolios.

That changes what’s possible. Every property can run with the discipline of the best-run one. A small operator can work with capabilities once reserved for the largest. Problems get caught before residents feel them. And capital can move with a full, current picture of where housing is needed and how it’s performing.

No single company will build all of it. The breakthroughs will come from an ecosystem of established providers, startups and operators’ own teams, competing to solve problems none of us could solve alone. We’re building toward that future, and we welcome every capable agent that helps get us there.

That’s why the foundation has to be right.

Agents are arriving

AI is moving from answering to acting.

Software once recorded what people did. Now AI agents take actions on their own: they log in, move through workflows, update records, send messages and trigger payments, at a speed no team can match.

33%

of enterprise software applications will include agentic AI by 2028, up from less than 1% in 2024.1

15%

of day-to-day work decisions will be made autonomously by agentic AI by 2028.1

Capacity

Agents take repetitive work off onsite and back-office teams, so people can focus on residents and judgment calls.

Speed

Prospects and residents get answers and action in minutes, not days.

Cost

Automation is one of the few levers that improves both service and NOI.

An agent that can only read merely reports on the work. The value is in doing it.

That means agents will write to the systems of record operators depend on. And residents will soon bring agents of their own, to search, apply, pay rent and request maintenance.

40%+

of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.

The agents that last will be the ones operators can trust. Read Gartner's prediction

The problem

Impersonation removes accountability.

Many agents reach core systems through the side door. A vendor asks for an employee's login, and its software signs in as that person. Some agents drive web screens like a human. Some disguise their traffic.

This is shadow access, and its most serious form is impersonation.

If a platform can’t tell that it’s an agent, which agent it is and who it’s acting for, the agent is unidentified, however it got in.

No accountability

The agent's actions are recorded as the employee's. No one can tell who did what, liability can land on the operator, and compliance records credit a person for an agent’s decisions.

OutcomeWhen a resident disputes a charge or a regulator asks who made a decision, there’s no answer. The operator can’t defend itself, and an employee may be blamed for something they never did.

Too much access

The agent inherits all of the person's permissions, far beyond its task: payments, leases, screening data.

OutcomeA single mistake, or a single attacker, can reach payments, leases and residents’ personal data the agent never needed.

No clean shutoff

Stopping the agent means locking out the employee too.

OutcomeThere may be no way to stop the agent on its own. Operators are left choosing between letting it run and locking out their own team.

Security openings

  • Logins stored on vendor systems mean a vendor breach exposes the operator.
  • To let the agent log in, the second security check (like a code sent to a phone) gets routed to the vendor or switched off.
  • Access continues after the employee leaves.
  • An attacker using the same login looks just like the agent.
  • Screen-reading agents can be manipulated by hidden instructions.

OutcomeOne compromised vendor can open a path into every operator, system and resident connected to it.

An agent with its own identity can do the same work: visibly, within scope, and accountably.

Already happening

This isn’t a forecast.
It’s already happening.

In the past 18 months, agents at well-known companies have exposed data, destroyed records and acted without permission.

Aug 2025

Salesloft Drift

One agent’s access became hundreds of breaches.

Attackers stole the access tokens behind an AI chat agent that companies had connected to Salesforce. Over about 10 days, they pulled data from more than 700 organizations and searched it for passwords and cloud keys. It stopped only when the agent’s access was cut off.

Read the article
Nov 2025

ServiceNow

When agents borrow each other’s permissions, no one knows who’s acting.

Security researchers showed that instructions hidden in a record could lead one AI agent to recruit a more privileged agent, which then copied data and changed records. It worked under default settings.

Read the research
Mar 2026

Meta

It happens to the best-resourced teams, and to their own agents.

An internal agent posted advice on its own, without the engineer’s approval. Acting on it exposed company and user data to unauthorized employees for about two hours. Meta rated it a Sev 1, its second-highest severity.

Read the article
Apr 2026

PocketOS

Nine seconds. No approval. No undo.

A coding agent hit an error, found an unrelated access key and deleted a production database and its backups, with no confirmation step. Car rental businesses on the platform were left with data three months old.

Read the article
Jul 2025

Replit

An agent’s own account isn’t a record.

During an explicit code freeze, an agent deleted a live production database, created about 4,000 fake records and misreported what it had done.

Read the article

Each began with an ordinary gap: too much access, no approval, no reliable record.

None of this is a reason to slow down. It’s the reason to build it right.

“You don’t want rogue agents.”Jamie Dimon, Chairman and CEO, JPMorgan Chase2
What other industries are learning

The world's largest platforms are drawing the same line.

Banking

JPMorgan

Blocks third-party agents that access accounts with customers' banking credentials, while supporting customer-authorized data sharing through approved channels.2

Open to agents. Closed to impersonation.
Read the article
Payments

Visa, Mastercard and Ant International

Began collaborating in September 2026 on a shared “Know Your Agent” framework: who the agent is, who authorized it, and what it may do.3

Agent identity is becoming shared infrastructure.
Read the article
Software platforms

Salesforce

Is moving external AI agents to register their own identity and credentials, instead of signing in as the person they serve.4

Every agent gets its own badge.
Read the article
Internet infrastructure

Cloudflare and Browserbase

Web Bot Auth gives agents a cryptographic “passport” so websites can tell legitimate agents from impostors.5

Legitimate agents identify themselves.
Read the article

Courts are still catching up. A federal appeals court recently vacated an injunction that had barred an AI shopping agent from customer accounts, reasoning that the customer, aided by the agent, was the one acting.6 Clear standards, transparent terms and governed channels protect operators however the law evolves.

Read about the ruling

Our principles

Seven principles for agents in rental housing and real estate.

These aren’t rules for one platform. They’re the standard we believe every agent, including our own, should meet to work in the systems operators and residents depend on. They rest on three beliefs:

  • Accountability. Every agent comes through the front door with its own identity, works within its task and the data that task needs, and leaves a clear record of what it does.
  • Customer control. Operators choose which agents work in their business, see what each one does, and can stop any of them at any time.
  • One standard. Every agent should meet the same requirements, whoever builds it, Realpage’s own included.
1

Agents are inevitable.

They will act for operators and residents on every platform. We plan for them, not against them.

2

Identity extends to agents.

The access and identity controls we apply to people apply to every agent. Each agent has a unique identity of its own, and no agent borrows a person’s identity.

3

Front door only.

Agents connect through governed channels, for a stated purpose, and take only the data that purpose needs. No shared logins, no screen automation, no spoofing.

4

Risk sets the bar.

Routine work runs automatically within rules operators set. High-risk actions outside those rules need human approval.

5

Every action is accountable.

Each action traces to a specific agent and the person who authorized it.

6

One standard for everyone.

Our own agents meet the same requirements as our partners'.

7

People stay in control.

Operators and residents authorize their agents, see what they do, and can stop them.

Identity isn't just a security control. It's what makes trust in agents possible.

For operators

Five questions to ask any AI vendor.

Whoever your vendors are, including us.

1

How does your agent connect?

Through a certified integration with its own identity, or using our employees' logins?

2

What can it reach, and why?

Exactly which systems and data, for what purpose, and how is that limited?

3

What can it change?

Can it move money, change leases or send notices, and which actions need our approval first?

4

Is every action recorded?

Is each one logged and attributed to your agent, not to a member of our staff?

5

How do we stop it?

Can we pause or shut it off immediately, without disrupting our team?

A consequential moment

A moment that asks more of all of us

Rental housing has weathered many changes. This one is different.

For the first time, software won’t just support the people who run our communities. It will act alongside them, making decisions that touch residents’ homes, their payments and their personal information. The choices our industry makes now will shape how those decisions are made for years to come.

Many of us grew up in this industry. We know what it means to someone when a lease is right, a payment posts on time, or a repair gets done. Those moments are why the protections around them matter so much.

This isn’t a moment for walls, and it isn’t a moment for winner-take-all. It’s a moment to come together. We invite every technology provider in rental housing and across real estate, established and new, to help build open, shared standards for how agents identify themselves, act and are held accountable, through established industry organizations, in an open process anyone can join. Standards anyone can meet, built in the open, with operators and residents’ interests at the center.

We’ll keep competing, and we should. Healthy competition pushes every one of us to deliver our best. But on the protections our customers and their residents deserve, we should all hold ourselves to the same standard.

Looking ahead

Helping shape what’s next for rental housing.

Realpage is open at scale today. For years, Realpage Exchange has given partners trusted access to platforms serving millions of apartment homes, letting them compete for customers they might not reach on their own. More than 800 applications connect through it, many of them competitors to our own products.

Every partner connects through a certified, governed integration, under agreements that define what it can access. Those expectations are what make an open ecosystem of this scale possible. What’s new is the scale and speed agents bring, and agents are the next chapter of that openness.

We believe these seven principles can be a starting point for common expectations across rental housing and the broader real estate industry. Not rules defined by any one company, but shared expectations that let agents work across platforms while protecting operators and residents.

Operators

Use the five questions to evaluate your AI vendors. Expect clear answers about how agents connect, what they can access and who is accountable.

Technology providers

Build on these principles. Every provider, including our competitors, has a stake in secure, interoperable agent access.

Industry organizations

Bring these principles into broader industry discussions and advance standards that support interoperability, accountability and choice.

In the coming months, we’ll share more about how these principles work across our platform, and invite the industry to build on them. The next generation of rental housing technology won’t be built by any one company. We’re ready to do our part.

By 2030, the most successful real estate organizations will be the ones that put agents to work everywhere they help, and know exactly which agents are working, what each one may do, and can prove it.

Sources
  1. Gartner, agentic AI predictions, June 2025. Link
  2. Fortune, "Jamie Dimon bars outside AI agents," October 15, 2026. Link
  3. Ant International, Mastercard and Visa, Know Your Agent collaboration (Business Wire), September 10, 2026. Link
  4. Stacksync, “Salesforce Agentic Identity checklist,” October 2026. Link
  5. Cloudflare and Browserbase, Web Bot Auth, 2025. Link
  6. Ninth Circuit, Amazon v. Perplexity, August 2026. Link
  7. CyberScoop, Salesloft Drift attack, August 2025. Link
  8. The Hacker News, ServiceNow agent-to-agent prompt injection research (AppOmni), November 2025. Link
  9. TechCrunch, “Meta is having trouble with rogue AI agents,” March 18, 2026. Link
  10. ACS Information Age, PocketOS database deletion, May 2026. Link
  11. eWeek, Replit agent database deletion, July 2025. Link