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
- A listing description that wrote itself into trouble
- AI in real estate, desk by desk
- Real estate AI tools, sorted by the job they do
- AI for real estate agents and brokers
- Property management and leasing
- Valuation and automated valuation models
- Lending: reasons a person can read
- Fair housing rules that reach AI-written ads and ad delivery
- Listing data: what an AI tool reads
- Building custom AI tools for a real estate firm
- A risk ladder for real estate and AI
- What AI tools for real estate cost to build
- 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.
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.
| Desk | What AI can do there | Who signs off | The rule nearby |
|---|---|---|---|
| Lead desk | Answer website and portal inquiries at any hour, qualify and route | The agent who takes the lead | Fair Housing Act, 42 U.S.C. 3604 |
| Listing desk | Draft descriptions, tag photos, check fields for gaps | The listing agent | 42 U.S.C. 3604(c) and 24 CFR 100.75 on statements in ads |
| Showings | Offer times, confirm, reschedule | The showing agent | 42 U.S.C. 3604(d) on stating availability |
| Leasing office | Answer questions, collect applications, summarize files | The leasing manager | 24 CFR 100.500 on discriminatory effect |
| Maintenance | Sort requests, set priority, dispatch vendors | The property manager | Lease terms and local codes |
| Pricing and valuation | Suggest list prices and rents from comparable data | The broker or appraiser | 12 CFR 1026.42(i) when an AVM values loan collateral |
| Loan desk | Pre-fill applications, flag missing documents, model risk | The loan officer and credit team | 12 CFR 1002.9(b)(2): specific reasons for adverse action |
| Transactions | Track deadlines, read documents, build checklists | The transaction coordinator | Contract dates and state rules |
| Marketing | Write ads, pick audiences, schedule posts | The marketing lead | 24 CFR 100.75(c)(3) on selecting media |
| Investor desk | Model returns, compare markets, summarize rent rolls | The analyst | The 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.
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.
| Tool type | What it does | What it reads | Check before buying or building |
|---|---|---|---|
| Listing writer | Drafts descriptions and ad copy | Listing fields and photos | Drafts describe the property, never the people who should live there |
| Lead assistant | Answers and qualifies inquiries | Listings, policies, calendar | Same answers and same follow up for every inquirer |
| Showing scheduler | Books and confirms visits | Agent calendars, lockbox data | Who sees which times, and why |
| Lease and document reader | Summarizes leases and contracts | Uploaded documents | Summaries checked against the source before anyone relies on them |
| Screening model | Scores rental applications | Application and report data | Discriminatory effect tested, with a legitimate reason for each factor |
| Pricing model | Suggests list prices and rents | Sold and leased comparables | Confidence, data quality and regular sample tests |
| Underwriting model | Supports credit decisions | Application and credit data | Specific principal reasons can be given for each denial |
| Ad audience tool | Chooses who sees an ad | Platform audience settings | No setting that leaves out people in a protected class |
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.
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.
| Rule | What it says, in short | What 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 unlawful | Every 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 information | Ad delivery and audience settings get the same review as the words |
| 24 CFR 100.500 | A practice can violate the Act through its discriminatory effect even without discriminatory intent; a justification must rest on evidence, not speculation | Screening 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 insufficient | An 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 models | A valuation model needs confidence measures, data controls, testing and a fair lending check |
| HUD’s reporting page | Property owners, managers, developers, agents, lenders, associations and insurers can all have allegations filed against them | Every 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.
- Pick one desk and one job with a clear result.
- Map the data and who may see it.
- Decide who signs off on each output.
- Build the tool grounded in the firm’s own records.
- Test it against real and tricky cases.
- 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 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.
- 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.
| Build | Typical low to published high | Notes |
|---|---|---|
| Single workflow automation | $2,000 to $8,000 | One job on the firm’s existing systems |
| Website chatbot with integrations | $10,000 to $30,000 | CRM and calendar connections included |
| AI proof of concept | $15,000 to $40,000 | Tests one idea on real data |
| AI feature in an existing app | $15,000 to $60,000 | Adds AI to a portal or CRM already in use |
| AI MVP | $40,000 to $120,000 | A new product with AI at its center |
| Tenant or owner portal | $50,000 to $120,000 | Often the home for leasing tools |
| Agent CRM | $80,000 to $180,000 | Lead routing and follow up built in |
| US development team, hourly | $100 to $200 an hour | Agency or contract bill rate |
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
Getting found in search
AI, AEO and what is changing
Paid media and lead generation
Websites and design
Choosing and working with an agency
Software and app development
Website design by industry and type
Web development, platforms and hosting
Social, content and brand
By industry and by situation
Frequently asked questions
What does AI in real estate actually do day to day?
Which real estate AI tools are worth trying first?
Is AI for real estate agents mainly about listing descriptions?
Can an AI tool post a listing description by itself?
Who is answerable when an AI-written ad shows a preference?
Can an AI tool decide which renters see a housing ad?
Can a tenant screening model be unfair without anyone meaning it?
What must a lender tell an applicant an AI model turned down?
What are the quality control standards for automated valuation models?
Do the valuation standards cover a brokerage pricing tool?
What listing data can an AI tool read?
What is the RESO Web API?
Should a brokerage buy an AI tool or build one?
How much does a custom AI tool for real estate cost?
Does an AI chatbot replace a leasing agent?
How should a firm test an AI tool before clients see it?
Is there a federal framework for managing AI risk?
Where can someone report housing discrimination?
Which protected classes does the Fair Housing Act list?
What should stay with a person in a firm using AI?
Can Progression build AI tools for our real estate firm?
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.
