Commercial real estate has never been more willing to try artificial intelligence, or less sure what it is getting for the money. JLL’s 2025 research on the industry found that roughly 90 percent of real estate companies are piloting AI, yet only about 5 percent say they have achieved all of their AI goals. Morgan Stanley estimates that AI could unlock $34 billion in operating efficiencies across real estate by 2030. Outside the industry, the organizational response has been fast: IBM’s May 2026 CEO study found that 76 percent of surveyed organizations now have a Chief AI Officer, up from 26 percent a year earlier.
Most CRE firms live in the gap between those numbers. They subscribe to a lease abstraction tool here, an underwriting assistant there and a marketing writer somewhere else, and they hope the pieces add up to a strategy. BestCRE tracks 593 commercial real estate AI tools and has scored 450 of them on the 9AI Framework, which gives us an unusually clear view of what happens to those subscriptions over time. In our October 2026 audit, 20 of the 450 reviewed tools had shut down, been folded into another company’s product, been bought by a larger company, or changed their name. That is about one in every 22 tools, and it does not count the 19 vendors that simply moved to a new web address.
The usual conclusion is that buyers need better vendor due diligence. That is true, and it is incomplete. The deeper lesson is that most CRE AI products are now thin layers over the same handful of foundation models, while the layer that actually compounds in value, meaning a firm’s deal history, underwriting logic, documents and the judgment of its people, is the one firms keep handing to someone else. The best CRE AI strategy in 2026 is to put a named owner in charge of AI, run your own agents on your own data with models you can swap, buy only what you cannot reasonably build, and treat the best of what you build as a product you might one day sell.
This analysis sits inside BestCRE’s coverage of AI across the 20 CRE sectors. It draws on the full Best CRE AI Tools directory and the 9AI scores behind it, and it connects to our category coverage of CRE Workflow and Automation, CRE Underwriting and Deal Analysis and CRE Market Analytics and Data.
The CRE AI Market Is Already Consolidating, and BestCRE’s Own Database Shows It
The 20 changes in our audit fall into four groups. Six tools appear to have shut down: their websites no longer load, and we could not find an active product page anywhere in search results. They included a lease abstraction tool built on GPT-4 summaries, a tour book generator for brokers, a redevelopment site finder and a portfolio analysis platform for investors. Three more were folded into other products. Autodesk moved PlanGrid into Autodesk Build, Hootsuite absorbed Heyday into its engagement tools, and the website of a zoning research copilot now redirects to an AI meeting notes product. Six are now owned by larger companies: Cherre by RealPage, Datagrid by Procore, Matterport by CoStar Group, Evisort by Workday, Enodo by Walker and Dunlop, and ProspectNow by Buildout, which now sells it as Prospect by Buildout. The last five operate under new names, including Groundbreaker, which became Janover Connect, and Buxton, which now operates under the Audiense name.
The deal flow beyond our list points the same way. Over the past year, trade coverage and company announcements have tracked CoStar closing an $800 million acquisition of Zonda, MSCI agreeing to buy the climate risk firm First Street, Clear Capital acquiring the property image AI company Restb.ai, Venn acquiring the multifamily leasing AI Zuma, MRI Software acquiring PropTech Labs, and Deepki buying both EVORA Global and Metry. The pattern is consistent: large platforms are buying AI features rather than building them, and small point solutions are either acquired, absorbed or left behind.
An acquisition is not a failure, and some of these products are better funded today than they were a year ago. For the firm paying the subscription, though, every one of these events changes something that was supposed to be stable: the price, the roadmap, the data terms, the integrations, or the product’s existence. Each subscription is a dependency on a roadmap you do not control. When the tool goes away, the prompts, the workflow design and sometimes the data go with it.
Most CRE AI Products Are a Model, a Prompt, a Connector and a Screen
Open up the typical CRE AI product and you find four layers. At the bottom is a foundation model licensed from OpenAI, Anthropic, Google or an open-weight provider. On top of it sits a set of instructions tuned to one task, such as extracting lease terms or drafting an offering memo summary. Around that is a set of connectors that pull in documents and data, and on top of everything is a review screen where a person checks the output. Two years ago, building each of those layers was real engineering work. Today the models are rented by the month, the instructions are text, connector standards such as the Model Context Protocol make it routine to plug a model into email, file storage, spreadsheets and property systems, and agent tools from the major labs can generate a usable review screen in an afternoon.
Some vendors still have durable advantages, and our scores reflect them. Proprietary data networks are hard to copy: CompStak and CoStar earn their places because of the comps behind them, not the chat window in front. Deep integration into systems of record such as Yardi, MRI and Procore is hard to replicate. Regulated workflows, enterprise support and distribution all matter. Point solutions without any of those advantages are the ones most exposed, and they account for most of the shutdowns in our audit.
The implication for buyers is simple. If a product’s value lives mainly in its instructions and its connectors, a firm with an AI owner and access to the same models can rebuild it, and can add something the vendor never had: the firm’s own data, criteria and history.
A Discontinued Tool Can Now Be Rebuilt In-House in Weeks, Not Quarters
Take lease abstraction, the job of one of the tools that went dark. An in-house version starts with the leases already sitting in your shared drive or property management system. A frontier model with a long context window reads each lease and fills in your firm’s own abstract template rather than a vendor’s generic one. A second pass compares each clause against your playbook, so that an unusual co-tenancy provision, a radius restriction or a below-market renewal option is flagged for an analyst. The analyst approves or corrects each abstract in a simple review queue, and the approved fields flow into your rent roll, your Excel models or your property system. For a mid-sized portfolio, a first production version is typically a two to three week project in our experience, and the model costs are small next to the analyst hours it saves. The category leaders, such as Prophia, still make sense for large portfolios that want a managed service, but the gap between buying and building has narrowed sharply.
Redevelopment site screening, the job of another discontinued tool, follows the same pattern. An agent pulls parcel records, zoning designations, recent sales and permit activity for the markets you care about, scores each site against the rubric your investment committee already uses, and posts a ranked shortlist with its reasoning every Monday. The data comes from public sources and subscriptions most acquisition teams already pay for. What makes the in-house version better than the vendor’s is the rubric, because it encodes your firm’s view of what a good site is.
Broker tour books, the job of a third tool that disappeared, may be the easiest of all. An agent reads the listings selected in your CRM, pulls photos, floor plans and availability, assembles them into your firm’s branded template and hands the draft to the broker for edits. None of these builds requires a software team. Each requires someone who owns the workflow, knows what good output looks like, and has the authority to put the result into daily use. The scarce ingredient was never the vendor’s code. It was always the firm’s own criteria.
The Chief AI Officer Is the Missing Seat at Most CRE Firms
The jump in Chief AI Officer appointments that IBM measured, from 26 percent of organizations to 76 percent in a single year, reflects a lesson most industries learned quickly: AI programs without a named owner become collections of pilots. That is a fair description of where much of commercial real estate sits, and it goes a long way toward explaining why JLL found that so few firms have hit all of their AI goals. Pilots are cheap to start and easy to abandon. Production systems need someone accountable for them.
In a CRE firm, the Chief AI Officer role has six parts. The first is owning the model portfolio and its cost, including which models the firm uses, on which subscriptions, and with what spending limits. The second is owning the harness and connectors that let agents reach the firm’s documents, inboxes, spreadsheets and property systems. The third is choosing the first workflows by the hours they will save, not by how impressive the demo looks. The fourth is setting evaluation and sign-off rules so that every agent output a lender, investor or tenant might see is checked by a person until the error rate earns more trust. The fifth is governance over tenant data, lender data and, in multifamily, fair housing exposure. The sixth is making vendor decisions with a clear build, buy or partner test, so that subscriptions are added on purpose rather than by accident.
The title matters less than the ownership. A large owner or fund manager may hire a full-time leader. A regional operator or brokerage might give the role to a partner who already understands both deals and technology, or bring in a fractional Chief AI Officer for a few days a month. What does not work is leaving AI to whoever happens to be curious, or delegating it entirely to an outside IT firm that has never underwritten a deal.
Own the Harness, Rent the Models
An agent harness is the software layer that lets a model do real work: read files, use tools, follow a written procedure, remember context between steps, and log what it did so a person can check it. The major labs now ship capable harnesses with their own products, open-source orchestration frameworks are mature, and workflow tools such as n8n can connect agents to the rest of a firm’s systems. The strategic point is that the harness, the procedures and the connectors are the parts a firm should own, while the models underneath should stay interchangeable.
Model leadership changes every few months. A workflow written once and run on Claude, ChatGPT, Gemini or an open-weight model, depending on price and quality that quarter, protects a firm from the same vendor risk our audit documents. It also gives the firm real control over cost. Routine extraction and classification can run on cheaper models, heavier analysis can go to the strongest ones, and flat-rate subscriptions can carry most of the load instead of open-ended metered billing. Just as important, the documents and the outputs stay inside the firm’s own environment, which is where lenders, investors and auditors expect them to be.
Buy the Systems of Record, Build the Judgment Layer, Partner on Products
A practical CRE AI strategy sorts every workflow into one of three groups. Buy the systems of record and the data networks: accounting and property management platforms, construction management, market data, comps, foot traffic and the large integrated suites. These products carry network effects, compliance burdens and integration depth that no single firm should rebuild, and many of them now include capable AI features of their own.
Build the judgment layer, which is everything that encodes how your firm decides. Deal screening against your buy box, underwriting assumptions, investment committee memos, lender packages, buyer targeting, broker opinions of value, investor reporting narratives, lease abstraction to your own schema, owner outreach based on your sourcing criteria and research on your target markets all belong here. These workflows are where generic tools are weakest, because they cannot see your history, and where in-house agents are strongest, because they can.
Partner when something you built solves a problem your peers share. That is the moment to consider licensing it, pairing it with a partner’s product, or spinning it out. The simplest test for any subscription is to ask what you would lose if the vendor disappeared tomorrow. If the answer is a commodity feature, buy the best one and move on. If the answer is your process, build it and own it.
The Same Playbook Applies to Capital Markets, Investment Sales and Investment Analysis
Nothing about this argument is limited to property operations software. The highest-value workflows in commercial real estate sit in capital markets, investment sales and investment analysis, and they are exactly the ones that depend on a firm’s own judgment and relationships. With the Mortgage Bankers Association reporting that 17 percent of outstanding commercial and multifamily mortgage balances are scheduled to mature in 2026, owners and lenders are working through an unusually heavy refinance calendar. An in-house agent can watch every loan in a portfolio, flag maturities and covenant pressure months ahead, assemble a lender package from the latest rent roll and operating statement, and match each request against the appetite of the lenders the firm actually works with.
In investment sales, the same approach turns ownership records, recent trades and a brokerage team’s own relationship history into a ranked buyer list, drafts the first version of a broker opinion of value and an offering memorandum, and identifies 1031 exchange buyers whose timelines line up with a listing. On the acquisition side, it screens on-market and off-market opportunities against a buy box every week and puts only the fits in front of the team.
Investment analysis ties it together. Agents can pull rent rolls, operating statements and lease terms out of offering memorandums, run them through the firm’s own underwriting model and assumptions, stress test debt and exit scenarios, and draft the investment committee memo. None of these is a product a vendor can sell off the shelf, because each is only as good as the criteria, relationships and history behind it. That is precisely why they belong in-house.
BestCRE Runs Its Own Research on an In-House Agent Stack
We hold ourselves to the same advice. The database behind this article is maintained by agents we built and run ourselves. A weekly scout reads the CRE AI newsletters we subscribe to, searches for new launches, funding rounds and acquisitions, confirms each company’s official website, and adds qualified tools to our tracking sheet. A website check tests every vendor address on the list and flags the ones that stop responding or start redirecting somewhere else. A nightly audit checks hundreds of pages for scoring consistency, broken or outdated links, formatting problems and internal linking, and repairs what it can. The October audit described above, including the 20 status changes and the 19 moved web addresses, came from that stack.
None of it required a software vendor or a large engineering team. It runs mostly on flat-rate AI subscriptions, it uses models from more than one lab, and every rule it follows is written down in plain language that a person can read and change. That is the practical meaning of owning the harness: when a better model arrives, we switch it in, and when a vendor disappears, nothing we depend on disappears with it.
Vendors Still Win Where Data, Compliance or Scale Is the Product
An honest build strategy also knows where to stop. Vendors win where the product is the data network, as with comps, foot traffic, ownership records and market analytics. They win where compliance is the product, as in payments, tenant screening and regulated reporting. They win where the workflow is a commodity that every firm does the same way, such as meeting notes, transcription, scheduling and routine email drafting, and they win where a firm simply needs a supported product with a service level behind it.
In-house builds fail in predictable ways as well. They fail when nobody owns them after the first demo, when the underlying data is scattered across personal inboxes and desktop spreadsheets, when there is no evaluation step and errors reach investors, and when the build is designed to impress rather than to be used every day. Each of those failure modes is a management problem, not a technology problem, which is why the Chief AI Officer seat matters more than any individual tool.
Your Internal Build Could Be the Next CRE AI Company
Many of the tools in our directory started as internal projects at operating firms that realized their peers had the same problem. That path is more open now than it has ever been, because the cost of building has fallen while the value of real operating data and distribution has not. A firm that builds a strong underwriting assistant, a lease abstraction workflow tuned to a property type, or a sourcing agent for a niche asset class may be holding the start of a product.
There are three common ways to structure that opportunity. In a license arrangement, the firm keeps using the tool at no cost while a partner productizes and sells it, and the firm shares in the revenue. In a data partnership, the firm’s proprietary data powers a partner’s product under clear terms, and both sides share the economics. In a spin-out, a new company is formed with the firm as anchor customer and shareholder. The ideas behind several discontinued tools in our audit also remain valid, because the need did not go away when the company did, and a firm with distribution and data can often succeed where a standalone startup could not.
If you want a second opinion, whether on a financing, a sale, an acquisition or what to build, buy or partner on in AI, tell BestCRE Advisory what you are working on. It takes about three minutes, and you will see a recommended next step as soon as you submit.
Rent the Models, Own the Judgment
CRE AI vendors will keep arriving, raising money, merging and disappearing, and that churn is a sign of a healthy market rather than a failing one. The firms that come out ahead will not be the ones that picked the luckiest startup. They will be the ones that put a named owner in charge, kept their workflows and data in their own hands, and treated models and point solutions as replaceable parts.
The question to carry into your next software renewal is not whether the tool is good. It is whether the thing that makes it valuable belongs to the vendor or to you.
About BestCRE
BestCRE is the independent authority on commercial real estate AI, analysis and investment intelligence. We score CRE AI tools on the nine-dimension 9AI Framework, track markets and capital across the 20 CRE sectors, and publish research for owners, investors, lenders, developers and brokers who want institutional-quality insight without the vendor spin.
Frequently Asked Questions
What is a Chief AI Officer in commercial real estate?
A Chief AI Officer in commercial real estate is the person accountable for how a firm uses artificial intelligence across deals, assets and operations. The role covers the firm’s choice of models and their cost, the agent harness and data connectors, the selection of workflows to automate, the review rules for anything an investor, lender or tenant might see, governance over sensitive data, and the decision to build, buy or partner on each tool. The role has spread quickly outside real estate: IBM’s May 2026 CEO study found that 76 percent of surveyed organizations now have one, up from 26 percent a year earlier. In CRE, the seat can be a full-time executive at a large owner or fund manager, a partner who already understands both deals and technology, or a fractional officer who works with the firm a few days a month. The title matters less than having one accountable owner.
How do AI agents replace CRE software subscriptions?
Many CRE AI subscriptions combine four parts: a licensed foundation model, task instructions, connectors to documents and data, and a review screen. An in-house AI agent combines the same parts, but the firm owns the instructions, connectors and outputs, and can switch the model underneath when prices or quality change. For a workflow such as lease abstraction, the agent reads each lease, fills in the firm’s own abstract template, flags unusual clauses against the firm’s playbook and sends the result to an analyst for approval. A first production version of a workflow like that is typically a two to three week project. Agents do not replace systems of record such as property management, accounting or market data platforms. They replace the thinner point solutions layered on top, which are also the products most exposed to shutdowns and acquisitions.
How many CRE AI tools have shut down or been acquired?
In BestCRE’s October 2026 audit of the 450 CRE AI tools we had reviewed, 20 had changed status: 6 appear to have shut down, 3 were folded into other companies’ products, 6 are now owned by larger companies, and 5 operate under new names. That is roughly one in every 22 tools, and another 19 vendors had moved to new web addresses. Consolidation extends well beyond our list. Over the past year, trade coverage and company announcements have tracked RealPage acquiring Cherre, Procore acquiring Datagrid, CoStar closing an $800 million acquisition of Zonda, MSCI agreeing to buy First Street, Clear Capital acquiring Restb.ai and Venn acquiring Zuma. BestCRE now tracks 593 tools and checks the status of every vendor on a recurring basis, keeping reviews of discontinued products online with a notice and links to active alternatives.
Should a CRE firm build or buy AI tools in 2026?
Most CRE firms should do both, deliberately. Buy systems of record and data networks, such as property management, accounting, construction management, comps and market analytics, because their value comes from network effects, compliance and integration depth that one firm cannot reproduce. Build the judgment layer, meaning deal screening against your buy box, underwriting assumptions, investment committee memos, investor reporting and lease abstraction to your own schema, because those workflows depend on your history and criteria. Partner when something you build would be valuable to peers. The simplest test is to ask what you would lose if a vendor disappeared tomorrow. JLL’s finding that only about 5 percent of real estate companies have achieved all of their AI goals suggests that buying pilots without an owner or a plan is the most common failure, regardless of which tools are chosen.
How can a CRE firm start building its own AI agents?
Start by naming one person who owns AI and giving that person a budget and the authority to change how work gets done. Standardize on one or two foundation models your team already uses, and set up an agent harness with connectors to your file storage, email and spreadsheets so agents can reach your real documents. Then choose two workflows with high hours and clear right answers, such as pulling rent rolls and operating statements from offering memos or abstracting leases to your template, and build them with a person approving every output at first. Most firms can put a first workflow into daily use within two to four weeks. Track the hours saved and the error rate, expand to the next workflow once the first is trusted, and review your subscriptions against what you can now do in-house. Firms that want help can start with BestCRE Advisory.
Continue reading: Best CRE AI Tools: the full 9AI directory, Prophia Review: the lease intelligence category leader, and CompStak Review: why proprietary comps still win.
