Skip to main content Scroll Top

AI in Real Estate: Where It Works for Agents, Brokers and Investors

Updated November 2026 · Written and maintained by the Progression Agency strategy team

AI in real estate can already draft a listing, answer a lead at midnight and summarize a lease. The harder question is which desk should hand which job to software, and which decisions stay with a person because a federal rule reaches them. Listing copy, lead replies and showing schedules carry little risk. Tenant screening, credit decisions and automated valuations sit under federal fair housing and lending rules, so a person signs off on each one. This page goes desk by desk through a brokerage, a property manager, a lender and an investor, and names the rule at each desk. Progression’s own developers build custom AI tools for real estate firms, and the team is based in New York.

On this page · 13 sections
  1. A listing description that wrote itself into trouble
  2. AI in real estate, desk by desk
  3. Real estate AI tools, sorted by the job they do
  4. AI for real estate agents and brokers
  5. Property management and leasing
  6. Valuation and automated valuation models
  7. Lending: reasons a person can read
  8. Fair housing rules that reach AI-written ads and ad delivery
  9. Listing data: what an AI tool reads
  10. Building custom AI tools for a real estate firm
  11. A risk ladder for real estate and AI
  12. What AI tools for real estate cost to build
  13. Questions to settle before an AI tool touches a client

Want to be named in AI answers instead? Being named when buyers ask ChatGPT for an agent is answer engine optimization, covered on our AEO for real estate agents page.

Examples marked “Illustrative example.” are hypothetical, and no real business, listing or person is shown.

A listing description that wrote itself into trouble

Picture a hypothetical property manager who lets an AI tool draft rental listings. The first draft for a two bedroom ends with “best for adults without children.” Nobody typed those words, and they were minutes from posting on three sites.

Two drawn boxes. The first holds an AI drafted rental description ending with best for adults without children, marked as a preference about familial status. The second holds the reviewed version describing the property's features instead.Two drawn boxes. The first holds an AI drafted rental description ending with best for adults without children, marked as a preference about familial status. The second holds the reviewed version describing the property's features instead.
Illustrative example. A hypothetical AI drafted rental listing before and after a person reviewed it.

The Fair Housing Act makes it unlawful to publish any notice, statement or advertisement for a dwelling that indicates a preference based on familial status or another protected class. A tool can write the sentence; the firm publishes it. That is the pattern across every desk below: AI does the first pass, and a named person owns what goes out.

  • The tool drafts.
  • A person reviews against the rule for that desk.
  • The firm keeps a record of what was published and who approved it.

AI in real estate, desk by desk

A real estate firm is a set of desks, each with its own job, its own data and its own rules. The table sets out what AI can do at each one, who should sign off, and the federal rule that sits nearest.

Ten desks in a real estate firm, the AI job at each, and the rule nearby
DeskWhat AI can do thereWho signs offThe rule nearby
Lead deskAnswer website and portal inquiries at any hour, qualify and routeThe agent who takes the leadFair Housing Act, 42 U.S.C. 3604
Listing deskDraft descriptions, tag photos, check fields for gapsThe listing agent42 U.S.C. 3604(c) and 24 CFR 100.75 on statements in ads
ShowingsOffer times, confirm, rescheduleThe showing agent42 U.S.C. 3604(d) on stating availability
Leasing officeAnswer questions, collect applications, summarize filesThe leasing manager24 CFR 100.500 on discriminatory effect
MaintenanceSort requests, set priority, dispatch vendorsThe property managerLease terms and local codes
Pricing and valuationSuggest list prices and rents from comparable dataThe broker or appraiser12 CFR 1026.42(i) when an AVM values loan collateral
Loan deskPre-fill applications, flag missing documents, model riskThe loan officer and credit team12 CFR 1002.9(b)(2): specific reasons for adverse action
TransactionsTrack deadlines, read documents, build checklistsThe transaction coordinatorContract dates and state rules
MarketingWrite ads, pick audiences, schedule postsThe marketing lead24 CFR 100.75(c)(3) on selecting media
Investor deskModel returns, compare markets, summarize rent rollsThe analystThe firm’s own investment policies

Which jobs suit AI first?

Jobs with a clear input, a clear finished result and a person who reads every output. Lead replies, listing drafts, showing schedules and lease summaries fit that test, and a mistake in any of them is caught before it reaches a client.

  • The input already lives in a system you control.
  • The output is short enough to check in a minute.
  • A wrong answer is embarrassing, not harmful.
  • The same job happens dozens of times a week.

Where a person stays in the loop

Anywhere the output decides something about a person: who gets a lease, who gets credit, what a home is worth for a loan, and who sees a housing ad. Those desks keep AI as an assistant to a decision a person makes and can explain.

7: Protected classes. Race, color, national origin, religion, sex, disability and familial status under the Fair Housing Act.
4: Kinds of discriminatory ad. Listed in HUD's advertising rule, including choosing media that leaves people out.
5: AVM quality standards. Confidence, data integrity, conflicts, sample testing and nondiscrimination.
1-800-669-9777: HUD's fair housing line. Where people report housing discrimination by phone.
2023: NIST AI RMF released. A voluntary framework for managing AI risk, published January 26.
2.0: RESO Data Dictionary. Keeps listing fields and pick lists consistent across tools.

Real estate AI tools, sorted by the job they do

Most real estate AI tools fall into four kinds of work: words, conversations, numbers and decisions. The table lists eight common types, what each reads, and the question to settle before it goes live.

Eight kinds of real estate AI tools and what to check
Tool typeWhat it doesWhat it readsCheck before buying or building
Listing writerDrafts descriptions and ad copyListing fields and photosDrafts describe the property, never the people who should live there
Lead assistantAnswers and qualifies inquiriesListings, policies, calendarSame answers and same follow up for every inquirer
Showing schedulerBooks and confirms visitsAgent calendars, lockbox dataWho sees which times, and why
Lease and document readerSummarizes leases and contractsUploaded documentsSummaries checked against the source before anyone relies on them
Screening modelScores rental applicationsApplication and report dataDiscriminatory effect tested, with a legitimate reason for each factor
Pricing modelSuggests list prices and rentsSold and leased comparablesConfidence, data quality and regular sample tests
Underwriting modelSupports credit decisionsApplication and credit dataSpecific principal reasons can be given for each denial
Ad audience toolChooses who sees an adPlatform audience settingsNo setting that leaves out people in a protected class
Four bands. Words lists listing descriptions, replies, lease summaries and ad copy. Conversations lists lead chat, showing scheduling, maintenance requests and after hours answers. Numbers lists pricing suggestions, rent comparisons, underwriting models and dashboards. Decisions lists tenant screening, credit decisions and mortgage valuations.Four bands. Words lists listing descriptions, replies, lease summaries and ad copy. Conversations lists lead chat, showing scheduling, maintenance requests and after hours answers. Numbers lists pricing suggestions, rent comparisons, underwriting models and dashboards. Decisions lists tenant screening, credit decisions and mortgage valuations.
Real estate AI tools sorted by the job they do, from words to decisions.

Off the shelf or custom?

An off the shelf tool suits a common job on common data, such as drafting listing copy. A custom build earns its cost when the tool must read the firm’s own systems, follow the firm’s own review steps, or keep a record the firm controls. Our real estate software development page covers the portals, CRMs and platforms such tools usually live inside.

  • Buy when the job is generic and the data is public.
  • Build when the data is yours and the review step is specific.
  • Either way, write down who signs off before launch.

AI for real estate agents and brokers

For an agent, AI for real estate mostly means time back: replies drafted while showing a house, a listing description ready for editing, a checklist built from the contract. For a broker, it means consistency across many agents and a record of what went out.

Lead replies that treat every inquirer the same

An assistant on the website can answer questions about a listing at any hour and offer showing times. Set it to give every inquirer the same information and the same follow up, and route the conversation to an agent once a person is ready to talk.

Listing copy from the listing record

A writer that works from the listing fields, rather than from guesses, produces fewer errors. Instruct it to describe rooms, features and location facts, never the kind of household the home suits, and have the listing agent approve each draft.

Checklists from contracts

A document reader can pull dates and contingencies from a signed contract into a transaction checklist. The coordinator checks each date against the document before anyone relies on it.

  • Draft replies, never send them unread at first.
  • Generate listing text from fields, then edit.
  • Pull contract dates into a checklist, then verify.
  • Keep a log of drafts and approvals.

Marketing for agents and brokerages, including paid channels and the fair housing limits on targeting, is set out on our real estate marketing agency page.

Property management and leasing

Leasing offices are where AI in real estate meets the most repeat questions, applications and maintenance requests, which makes them a natural fit for AI assistants and a sensitive place for AI decisions.

  • Answer questions about availability, pets, parking and fees from the firm’s own policy documents.
  • Collect applications and flag missing documents.
  • Sort maintenance requests by urgency and dispatch vendors.
  • Summarize a resident’s file for the manager before a renewal.

Can a screening model be unfair if nobody meant it to be?

Yes. HUD’s rule on discriminatory effect says liability under the Fair Housing Act can rest on a practice’s discriminatory effect even without discriminatory intent. A practice may still be lawful if it is necessary for a substantial, legitimate, nondiscriminatory interest that a less discriminatory practice could not serve, and that justification must be supported by evidence rather than speculation. For a screening model, that means testing outcomes across protected classes and keeping a written reason for every factor it weighs.

Valuation and automated valuation models

Pricing suggestions and automated valuations are the numbers side of artificial intelligence in real estate. When a model values a home that secures a mortgage, Regulation Z sets quality control standards for it.

A drawn card with five rows: confidence in estimates, protection against data manipulation, avoiding conflicts of interest, random sample testing and reviews, and compliance with nondiscrimination laws.A drawn card with five rows: confidence in estimates, protection against data manipulation, avoiding conflicts of interest, random sample testing and reviews, and compliance with nondiscrimination laws.
A drawn card listing the five quality control standards for automated valuation models.

The rule defines an automated valuation model as any computerized model used by mortgage originators and secondary market issuers to determine the value of a consumer’s principal dwelling securing a mortgage. The section was last amended in August 2024. A brokerage’s list price suggestion is a different use, but the same five checks make a sound build standard for any pricing model a firm relies on.

  • Report a confidence level with every estimate.
  • Lock down who can change the input data.
  • Keep the people who profit from a value away from the model.
  • Test a random sample against real sales on a schedule.
  • Check results for differences across neighborhoods and protected classes.

Lending: reasons a person can read

Lenders use AI to pre-fill applications, sort documents and model risk. When a model contributes to turning an applicant down, Regulation B still requires a statement of specific reasons.

The regulation says the statement must be specific and indicate the principal reasons for the adverse action, and that saying the applicant failed to achieve a qualifying score on the creditor’s scoring system is insufficient. Build any underwriting model so it can produce those reasons in plain words, and test that the reasons match what actually drove each result.

What should an underwriting model log?

The inputs it used, the version of the model, the output, the principal reasons it produced, and the person who made the final call. That record lets the lender explain any decision later.

Which desk is drowning in busywork?Tell us the desk and the systems it uses. You get a free proposal for the AI tool, the review step and the price.

Get a free proposalTell us about a project you’d like help with.

Fair housing rules that reach AI-written ads and ad delivery

Two HUD texts matter most for real estate and AI in marketing: the Fair Housing Act’s advertising section and HUD’s advertising regulation. The table states them briefly with what each means for a tool.

Federal rules nearest to AI in real estate, in short
RuleWhat it says, in shortWhat it means for an AI tool
42 U.S.C. 3604(c)Making, printing or publishing any notice, statement or advertisement for a dwelling that indicates a preference, limitation or discrimination based on a protected class is unlawfulEvery AI drafted listing and ad is read before it posts
24 CFR 100.75(c)Discriminatory ads include words, photos or symbols conveying who may live there, and selecting media or locations that deny parts of the market informationAd delivery and audience settings get the same review as the words
24 CFR 100.500A practice can violate the Act through its discriminatory effect even without discriminatory intent; a justification must rest on evidence, not speculationScreening and pricing models are tested for effect, and each factor has a documented reason
12 CFR 1002.9(b)(2)Adverse action notices give specific principal reasons; saying the applicant missed a score on the creditor’s scoring system is insufficientAn underwriting model must be able to say why it said no
12 CFR 1026.42(i)(3)Covered mortgage originators and secondary market issuers keep quality control standards for automated valuation modelsA valuation model needs confidence measures, data controls, testing and a fair lending check
HUD’s reporting pageProperty owners, managers, developers, agents, lenders, associations and insurers can all have allegations filed against themEvery desk in the firm, not only leasing, is in scope

Can an AI tool choose who sees a housing ad?

It can make the choice, and HUD’s advertising rule still applies to the result. The rule lists selecting media or locations for housing ads that deny particular segments of the housing market information because of a protected class as a discriminatory practice. Review audience and delivery settings with the same care as the words.

  • Describe the property, its features and its location facts.
  • Leave out who the home is best for.
  • Review audience settings before every campaign.
  • Keep a copy of each ad and its settings.

HUD lists the Fair Housing Act’s protected classes as race, color, national origin, religion, sex, disability and familial status, and takes reports of housing discrimination online or at 1-800-669-9777, as its reporting page explains.

Listing data: what an AI tool reads

Every use of AI in real estate is only as accurate as the data it reads. In residential real estate much of that data moves through MLS feeds built on standards from the Real Estate Standards Organization.

RESO describes its Web API as the modern way to transport real estate data, built on RESTful design and OData, with the industry moving away from the older RETS transport. Its Data Dictionary is the shared vocabulary: version 2.0 keeps the fields and pick lists in listing data consistent from the MLS to consumer websites, organized as resources such as Property, Member and Office, with fields and lookup values under each.

Why does the data dictionary matter for a custom build?

Because a tool that maps to standard fields can read listings from more than one feed without a rewrite, and its outputs line up with what agents see in the MLS. Each MLS sets its own access and display rules, so confirm them before a tool reads or republishes listing data.

  • Map every field the tool uses to the standard name.
  • Record which feed each listing came from.
  • Respect each MLS’s display and use rules.
  • Refresh data on the feed’s schedule, not by hand.

Building custom AI tools for a real estate firm

Progression’s own developers build custom AI tools for brokerages, property managers, lenders and investors: lead assistants, listing writers, document readers and AI features inside portals and CRMs. The stages below show how such a tool comes together.

A six stage flow: pick one desk and one job, map the data, decide the sign off, build and ground it, test against real cases, and launch with logging and review.A six stage flow: pick one desk and one job, map the data, decide the sign off, build and ground it, test against real cases, and launch with logging and review.
Six stages for building a custom AI tool, each with a person accountable.
  1. Pick one desk and one job with a clear result.
  2. Map the data and who may see it.
  3. Decide who signs off on each output.
  4. Build the tool grounded in the firm’s own records.
  5. Test it against real and tricky cases.
  6. Launch with logging and a review schedule.

See our pages on AI app development, AI automation and AI chatbot development for the wider work.

A drawn checklist with seven rows. Five are ticked: drafts describe the property, a person approves every listing, facts come from the listing record, fields map to the data, and a log keeps each draft. Two are crossed: ad audience settings reviewed and a monthly test set.A drawn checklist with seven rows. Five are ticked: drafts describe the property, a person approves every listing, facts come from the listing record, fields map to the data, and a log keeps each draft. Two are crossed: ad audience settings reviewed and a monthly test set.
Illustrative example. A hypothetical listing writer passes five of seven checks before launch.

A risk ladder for real estate and AI

Not every AI job carries the same weight. The ladder sorts them by how much a wrong answer could hurt a person, which decides how close a human stays.

Three rungs. Top: approve or deny a renter, approve or deny credit, value collateral for a loan, where a person decides. Middle: publish a listing, target or deliver ads, suggest a list price, route leads, where a person reviews. Bottom: draft an email, summarize a lease, book a showing, tag photos, where a person spot checks.Three rungs. Top: approve or deny a renter, approve or deny credit, value collateral for a loan, where a person decides. Middle: publish a listing, target or deliver ads, suggest a list price, route leads, where a person reviews. Bottom: draft an email, summarize a lease, book a showing, tag photos, where a person spot checks.
A three rung risk ladder for AI jobs, from spot checks to decisions a person makes.
  • Bottom rung: spot check a sample each week.
  • Middle rung: a person reviews every output before it goes out.
  • Top rung: a person decides, using the tool as an input, and can explain the decision.

What AI tools for real estate cost to build

Custom AI work is priced by the job, the data and the integrations. The bands below run from the typical low to the published high in the US market.

AI builds for real estate firms: published US price bands (market ranges only)
BuildTypical low to published highNotes
Single workflow automation$2,000 to $8,000One job on the firm’s existing systems
Website chatbot with integrations$10,000 to $30,000CRM and calendar connections included
AI proof of concept$15,000 to $40,000Tests one idea on real data
AI feature in an existing app$15,000 to $60,000Adds AI to a portal or CRM already in use
AI MVP$40,000 to $120,000A new product with AI at its center
Tenant or owner portal$50,000 to $120,000Often the home for leasing tools
Agent CRM$80,000 to $180,000Lead routing and follow up built in
US development team, hourly$100 to $200 an hourAgency or contract bill rate
Horizontal range bars for a single workflow automation, a website chatbot with integrations, an AI feature in an existing app, an AI MVP and an agent CRM.Horizontal range bars for a single workflow automation, a website chatbot with integrations, an AI feature in an existing app, an AI MVP and an agent CRM.
Published US price bands (market ranges only) for five kinds of AI build, typical low to published high.

Progression’s price for a build is set in the written proposal, after scope, and is subject to change. Our AI automation cost and app development cost guides break the bands down further.

What moves the price of a custom build?

The number of systems the tool reads, how much review and logging it needs, whether it makes or supports decisions about people, and how many desks use it. A listing writer on one feed sits at the low end; a screening or underwriting tool with testing and audit records sits higher.

Pricing depends on scope and is confirmed in a written proposal. Published prices are subject to change.

Questions to settle before an AI tool touches a client

Before any use of AI in real estate reaches a client, settle these in writing. The NIST AI Risk Management Framework, released on January 26, 2023 for voluntary use, and its generative AI profile from July 26, 2024 are useful references for the same questions.

  • Which desk owns the tool, and which person signs off?
  • What data does it read, and is that data current?
  • Does any output decide something about a person?
  • How are fair housing and lending rules checked?
  • What is logged, and for how long?
  • How often is it tested on tricky cases?
  • Who can switch it off?

What this page leaves out: the Progression Method, proprietary and reserved for clients. Reach out to learn more

Frequently asked questions

What does AI in real estate actually do day to day?
It drafts listings and replies, answers leads, books showings, summarizes leases and suggests prices. Decisions about renters, credit and loan collateral stay with people because federal rules reach them.
Which real estate AI tools are worth trying first?
Tools for words and conversations: listing drafts, lead replies, showing schedules and document summaries. A person can read every output, and a mistake is caught before it reaches a client.
Is AI for real estate agents mainly about listing descriptions?
Listing copy is one use among several. Agents also use AI for lead replies, showing schedules, contract checklists and market summaries.
Can an AI tool post a listing description by itself?
It can, but a person should approve each one first. The Fair Housing Act covers any statement in a housing ad, so a tool’s draft is read against that rule before it posts.
Who is answerable when an AI-written ad shows a preference?
The firm that publishes it. The Act makes it unlawful to make, print or publish a housing ad indicating a preference based on a protected class, whoever drafted the words.
Can an AI tool decide which renters see a housing ad?
It can make the choice, and HUD’s advertising rule covers the result. Selecting media or locations that deny parts of the market information because of a protected class is listed as discriminatory.
Can a tenant screening model be unfair without anyone meaning it?
Yes. HUD’s rule on discriminatory effect allows liability without discriminatory intent, unless the practice is justified by evidence and a less discriminatory one could not serve the same interest.
What must a lender tell an applicant an AI model turned down?
The specific principal reasons for the decision. Regulation B says a statement that the applicant missed a qualifying score on the creditor’s scoring system is insufficient.
What are the quality control standards for automated valuation models?
Five: a high level of confidence in estimates, protection against data manipulation, avoiding conflicts of interest, random sample testing and reviews, and compliance with nondiscrimination laws.
Do the valuation standards cover a brokerage pricing tool?
They are written for mortgage originators and secondary market issuers valuing loan collateral. A brokerage’s list price suggestion is a different use, though the same five checks make a sound standard for it.
What listing data can an AI tool read?
Whatever the firm’s MLS feeds and own systems provide, within each MLS’s access and display rules. Mapping to RESO Data Dictionary fields keeps that data consistent across tools.
What is the RESO Web API?
It is the Real Estate Standards Organization’s standard for moving real estate data, built on RESTful design and OData. The industry is moving to it from the older RETS transport.
Should a brokerage buy an AI tool or build one?
Buy for generic jobs on public data, such as drafting copy. Build when the tool must read the firm’s own systems, follow its review steps and keep records it controls.
How much does a custom AI tool for real estate cost?
Published US bands run from $2,000 to $8,000 for a single workflow automation, $10,000 to $30,000 for a chatbot with integrations and $40,000 to $120,000 for an AI MVP. Progression sets its price in the written proposal.
Does an AI chatbot replace a leasing agent?
It takes the repeat questions and the after hours inquiries. A leasing agent still handles tours, judgment calls and every decision about an applicant.
How should a firm test an AI tool before clients see it?
Run it on a set of real and tricky cases, read every output, and check outcomes across protected classes for anything that screens, prices or targets. Repeat the test on a schedule after launch.
Is there a federal framework for managing AI risk?
NIST’s AI Risk Management Framework, released January 26, 2023, is intended for voluntary use. NIST added a generative AI profile on July 26, 2024.
Where can someone report housing discrimination?
With HUD’s Office of Fair Housing and Equal Opportunity, online or by phone at 1-800-669-9777. HUD’s page lists owners, managers, developers, agents, lenders, associations and insurers among those who can be named.
Which protected classes does the Fair Housing Act list?
Race, color, national origin, religion, sex, disability and familial status, as HUD lists them.
What should stay with a person in a firm using AI?
Any decision about a renter, a borrower or a home’s value for a loan, and the final approval of every published ad. AI can prepare those decisions; a person makes and explains them.
Can Progression build AI tools for our real estate firm?
Yes. Progression’s own developers build custom AI tools for brokerages, property managers, lenders and investors, and the team is based in New York with on-site sessions available on request.

Which desk is drowning in busywork?Tell us the desk and the systems it uses. You get a free proposal for the AI tool, the review step and the price.

Get a free proposalTell us about a project you’d like help with.

Get a free marketing proposal

Tell us what you are trying to grow and we will come back with a plan, not a pitch deck. Same-day reply on weekdays.

Privacy Preferences
When you visit our website, it may store information through your browser from specific services, usually in form of cookies. Here you can change your privacy preferences. Please note that blocking some types of cookies may impact your experience on our website and the services we offer.
Contact Us