AI Real Estate Companies: 21 Defining the Industry

Why This Is a Map and Not a Ranking

We said in the guide to AI in real estate that we would not publish a ranked list of vendors, and that position has not changed. A ranking implies someone ran these products side by side on the same portfolio, in the same market, with the same team, and measured the outcome. Nobody has. The lists that pretend otherwise are ordered by funding, by affiliate commission, or by nothing at all.

What can be done honestly is a map. Grouping is a real claim: it says these three companies solve the same problem and therefore compete, and it lets you place the vendor who just emailed you into a category whose failure modes you already understand. That is more useful than a number out of ten, and it survives the next funding round.

• Every description here is what the company does, not how well it does it. Where a specific figure is quoted, it is attributed and dated, because these numbers move every quarter.
• Ownership is noted wherever it matters. Several entries on other people's lists are no longer independent companies, which changes the roadmap, the pricing and who you are actually signing with.
• Nothing here is a recommendation. We build automations for real estate teams; we are not a reseller of any of these products and there are no referral links on this page.
• Company facts were checked in July 2026. Proptech consolidates fast. Verify status before you sign anything, and treat any list without dates as fiction.

Valuation, Pricing and Market Data

This is the oldest AI in the industry and the least discussed as AI, because it predates the current wave by fifteen years. These are prediction models: they consume historical transactions and return an estimate. The category's failure mode is the confident number in a thin market.

Commercial Data and Lease Intelligence

Commercial real estate has the opposite problem to residential. The volume is low, the documents are long, and the value sits in extracting structure from PDFs nobody wants to read. These three companies attack different parts of that.

Notice that one of the three is explicitly infrastructure. That is the pattern in commercial: the model is rarely the hard part, and the project is almost always a data project wearing an AI budget. We cover that in more depth in the guide to AI tools for commercial real estate.

Conversational and Leasing AI

This is the category with the clearest return, because it fixes response time and response time decides who gets the deal. It is also the category where the same product is sold to residential brokerages and to multifamily operators with completely different economics.

If this is the category you are shopping in, the mechanics matter more than the logo: what it does when it does not know an answer, whether it writes to your CRM, and whether it discloses that it is an AI. We wrote the whole thing up in the guide to AI chatbots for real estate, and the chatbot service page is the build-it version.

Agent Platforms and Lead Engines

These companies sell a system of record with AI features layered on: a CRM, a website, ad management and follow-up in one subscription. The category's failure mode is not the AI. It is the switching cost.

Property Management and Revenue Operations

Property management is where AI has been deployed at the largest scale, because the work is repetitive, high-volume and measurable. It is also the only category on this page with an antitrust judgment attached to it, which is why it needs to be bought differently from the others.

The RealPage case is worth understanding even if you will never buy the product, because it settled a question the whole industry had been avoiding. An algorithm that prices your inventory using competitors' non-public data is not treated as clever software. It is treated as coordination. Separate class actions brought by renters were still working through the courts in 2026.

Listing Media and Computer Vision

Photographs are the most valuable unstructured data in real estate, and the only category here where the AI is looking at pixels rather than numbers or text. It is also the category with an explicit disclosure obligation.

The line that keeps firms out of trouble is simple and worth stating: enhance the lighting, disclose the staging, never move a wall. Anything that changes what a buyer believes is physically present is a misrepresentation regardless of which software produced it. The wider compliance picture, including the EU transparency rules that came into force in August 2026, is covered in the guide to generative AI in real estate.

Design, Feasibility and Development AI

The least visible category and arguably the one doing the most technically interesting work. These tools answer the question a developer used to pay an architect three weeks to answer: what can I actually build on this site, and does it pencil?

The Firms Building It In-House

Three of the biggest AI operations in real estate do not appear on vendor lists because they do not sell software. They are brokerages and service firms that decided their data was the product.

• JLL announced the acquisition of Skyline AI, a multifamily valuation and origination modelling company, back in August 2021, and has since built internal large-language-model tooling on top of its own research and transaction data. It is a vendor to its clients and a competitor to everyone in the CRE data category.
• CBRE has taken the partnership-and-platform route, wiring analytics into its service lines rather than productising them, which means the AI arrives inside an advisory engagement rather than as a licence.
• Compass built an agent-facing platform as its central thesis and spent years arguing that the technology, not the brand, was the reason to join. That claim is still contested, but the platform is real and it is not for sale.
• Opendoor remains the largest bet ever placed on a pricing model in residential real estate: the model does not advise on the price, it decides what the company pays. That is a materially different risk profile from any vendor here.

The reason this matters to a smaller firm: it tells you where the ceiling is. Every one of these organisations concluded that the differentiator was proprietary data plus workflow, not access to a model. That is available to a fifteen-agent brokerage too, at a fraction of the cost, because you already own transaction history nobody else has.

Who Quietly Left the List

Every list of this kind is a snapshot presented as a landscape. The instructive part is what has fallen off, because it shows which categories were businesses and which were features.

• Zillow Offers. The most heavily promoted algorithmic buying operation in residential real estate was wound down in late 2021 after the pricing model's errors met a moving market. The lesson was not that the model was bad, it was that a model taking balance-sheet risk fails differently from one producing a suggestion.
• Skyline AI. Headlined these lists for years and is now a capability inside JLL. Acquisition is the most common ending for a good real estate AI company, which is exactly why the ownership column matters when you sign a three-year contract.
• Standalone AI writing tools for agents. A wave of 2023-era listing-description generators has largely disappeared into features of the CRMs and portals. Anything that is one prompt wide gets absorbed by the platform that already holds the data.
• Independent point solutions in leasing. The multifamily category has consolidated hard, with leasing AI, CRM and call handling collapsing into bundles and partnerships. Buying a point solution in a consolidating category means budgeting for a migration you did not plan.

One Question Per Category, Asked in the First Call

Vendor evaluation goes wrong because buyers ask about features, which every vendor has rehearsed. These seven questions are chosen because a scripted demo cannot answer them and an honest company can.

One more that applies to every category: ask who does the implementation. A large share of failed real estate AI projects are not product failures. They are integration projects nobody scoped, discovered in month three, and the vendor's answer is usually a partner you have not met.

What a Small Team Should Do With This List

Here is the uncomfortable part. Read the caveats column again and count how many of these twenty-one companies would sell to a team of eight agents at a price that makes sense. It is a short list, and it is short for a rational reason: these products are built for the customers who can pay for enterprise sales.

That is not a reason to sit out. It is a reason to stop shopping and start measuring. Most small firms do not have a software gap, they have a response-time gap, a data-entry gap and a follow-up gap, all three of which are automation problems rather than platform problems.

1. 1Find the leak before the vendor does. Take last month's enquiries and measure the real gap between arrival and first human reply. That single number decides your priority far more reliably than any category on this page.
2. 2Fix the first reply. It is the cheapest intervention with the fastest payback, and it works whether or not you ever buy a platform.
3. 3Get your property data into one place. Every capability on this page — valuation, generation, extraction, leasing AI — performs in proportion to how clean and how central your own data is. This step is unglamorous and it is the whole game.
4. 4Wire the tools you already pay for. Your CRM, your calendar, your WhatsApp and your website can usually be connected into something that behaves like the platforms above, for a fraction of the licence cost. That is what CRM automation actually means in practice.
5. 5Only then buy a product. And buy it for one named job, with a number attached, and a date by which you will know whether it worked.

If you want the tool-level version of this rather than the company-level version, the two companion pieces are AI tools for real estate agents for residential teams and AI tools for commercial real estate for portfolio work. And if the question underneath all of this is how the models themselves work well enough to be trusted, that is machine learning in real estate.

Frequently Asked Questions

Conclusion

The twenty-one companies here are defining the industry in a narrower sense than the phrase suggests. They are not replacing anybody. They are absorbing specific tasks — valuing, extracting, answering, staging, laying out a site — and the ones that will still be independent in five years are those holding data or a workflow that cannot be re-created by pointing a general-purpose model at the problem.

For anyone reading this as a buyer, the useful takeaway is not a name. It is that the category tells you the failure mode, the ownership tells you the roadmap, and your own response-time number tells you whether you need any of it yet. Most firms discover, once they measure, that the expensive gap in their business was never on the shortlist.