How Brokers Can Use AI to Win More CRE Listings
AI tools compress the work that separates prepared brokers from their competition.
Winning a CRE listing mandate in 2026 depends more on how prepared a broker walks into the room than on how many years sit on a resume. The Mortgage Bankers Association projects commercial mortgage origination will rise meaningfully in 2026 from the prior year, and that volume increase pushes more deals into motion at once. More deals in motion means more pitches, more competing brokers in front of the same seller, and far less patience for a generic market deck. A seller comparing three brokers who all show up with similar comp sets and similar promises has little basis to choose on substance alone, so the choice tilts toward whoever demonstrates, in the room, that they already understand the asset, the debt markets, and the likely buyer pool. That gap between brokers who show up prepared and brokers who show up with boilerplate is not a matter of style or personality. It comes from a structural fact about the work itself: manual listing preparation, offering memorandum drafting, and capital markets research each eat up many hours of analyst time, hours that AI tools can now compress by a wide margin. Preparation has become a workflow problem, and workflow problems have workflow solutions, which is the premise the rest of this piece works through, section by section, from the listing copy a broker publishes in an afternoon to the capital markets argument that wins the mandate.
Compressing listing content from a day's work to under two hours
The clearest and lowest-risk place for a broker to start with AI is the gap between raw MLS data and a published listing, because the time savings there are large enough to change how many listings a single team can chase at once. Agentic AI deployments in brokerages now compress the MLS-to-publish cycle sharply: the AI agent drafts listing copy, social media variants, and email blasts, and it does this in a fraction of the time the process used to take. The output isn't limited to the listing description itself. The same MLS input feeds social posts and email blasts at once, so a broker who once needed half a day to build the marketing package for a single listing can now get through several listings before lunch.
None of this removes the broker from the loop, and it shouldn't. Fair Housing compliance and factual accuracy still require a human check before anything goes out, so the AI produces the draft and the broker makes the publish decision. That division of labor is the whole point: the agent handles repetitive drafting at speed, and the broker spends review time on judgment calls rather than first-draft assembly. If a seller is sizing up which broker to hire, a team that can turn an MLS entry into polished, compliant marketing within hours sends a clear signal about how it works under deadline pressure, and a competitor still writing listing copy by hand can't match that signal on the same timeline.
Listing copy is the visible layer of the pitch, the part a seller can glance at and judge in seconds. A seller trusts a broker to run the whole process based on the documentation underneath the listing copy, the offering memorandum and the market intelligence that back it up.
Offering Memorandum Production Time and Quality
If listing copy is what a seller sees first, the offering memorandum is what a seller's advisors actually read, and it is where a broker's preparation gets evaluated most closely. AI can cut OM production time from weeks of analyst work down to a matter of hours, and it can do so without giving up the polish that wins the mandate in the first place. A traditional OM, built from a property narrative, a market overview, a tenant summary, and a financial summary, takes a large number of analyst-hours to produce, and most of that work has to be rebuilt nearly from scratch for every new assignment because each property and market is different.
A tool like Claude changes that math by automating the sections that eat the most time: the property narrative, the market overview, and the tenant summary. The workflow itself is specific. An analyst feeds the rent roll, the T-12, lease abstracts, and seller due diligence materials into the model, then adds the firm's own recent OMs as references for voice and structure, and what comes out reads like a senior analyst's first draft. The generic-prose objection is real but solvable: uploading prior OMs as style references and explicitly prompting for that firm's voice produces output that reads as institutional, not templated.
None of this changes one rule that has to govern the work: the model handles structure and language, but every figure in the document needs a verified primary source behind it. An AI model can produce a cap rate that sounds entirely plausible and is simply wrong, and if that number ends up in a published OM, the broker's credibility is the thing on the line, not the model's. Framed correctly, that rule is the professional discipline that separates a broker who understands exactly where AI helps from one who treats it as a shortcut, and sellers can tell the difference. A broker who brings a draft-quality OM to the pitch meeting, instead of promising one weeks after winning the assignment, shows a level of readiness that a rival working at manual pace cannot match in the same window of time.
Building deal packages AI-assisted buyers can read
AI now has a hand in commercial real estate deals worth hundreds of billions of dollars, and a buyer's team puts it between itself and the offering memorandum as a reader, not as a tool a broker simply chooses to use. Buyers and their advisors increasingly run the first pass on a deal package through an AI tool before a human reads it closely, and that shift gives a structural advantage to brokers who control what the AI actually sees. That advantage has nothing to do with how well the OM is written in a literary sense.
It has to do with format. PDF-based deal packages frequently lose context when an AI system processes them: tables collapse, footnotes get separated from the figures they qualify, and image-based pages return nothing useful, all of which can produce inaccurate or incomplete answers when a buyer's AI tool reads the document. The Duxre AI Brief offers one early, concrete example of brokers addressing this directly: it ships free with every listing published to the Duxre marketplace, and a broker can create a free Ignite account, publish a listing, and generate the Brief from their own materials at no cost. The tool is evidence of where the market is heading, one way among others to solve the problem.
The underlying logic applies regardless of which tool a broker uses. A deal package built with clean data, reconciled figures, and broker-approved summaries lowers the odds that a buyer's AI misreads a number or a lease term and raises an objection rooted in a processing error rather than a real flaw in the deal. That makes OM quality a two-audience problem now. A broker has to satisfy the seller's advisors reading the document by hand and the buyer's AI reading it first, and the brokers who recognize that split early are building a pitch advantage that most competitors haven't noticed yet.
How AI-powered capital markets intelligence changes the buyer conversation
Documentation only carries a pitch so far. The broker who walks into a listing presentation already knowing which buyers are most likely to transact, and why, is having a fundamentally different conversation than the one who promises to "run a process" after getting selected. JLL's AI-powered buyer prediction platform shows what that looks like at institutional scale: it identifies a large share of capital markets buyers before they formally enter the market, drawing on data from millions of properties and tens of thousands of transactions a year. An independent broker won't match that scale, but the underlying approach, pairing asset data with debt market conditions to predict buyer behavior, is the same approach, just applied at a smaller size.
The pitch version of that intelligence sounds like this: a broker who can tell a seller, based on current debt markets and this asset's specific profile, which three buyer types are most likely to move and why the capital stack works for each of them, is making an argument grounded in market intelligence rather than one grounded in relationships.
Getting to that argument requires the underwriting work behind it to move fast, and AI agents now automate much of the extraction that used to slow it down: pulling data from rent rolls, T-12s, and lease abstracts feeds directly into the underwriting conversation, cutting the time analysts spend on manual data entry and compressing deal approval cycles in deployments where firms have measured the effect. The actual bottleneck in underwriting was never the financial model itself. It was getting real numbers into that model in the first place, and an AI underwriting copilot handles the extraction and pre-fills the model so the analyst's job shifts from retyping figures to reviewing and stress-testing them. Lev's platform builds its financial spreading and capital markets tools around that shift, so a broker can walk into a pitch with a financial narrative already grounded in verified numbers, not estimates assembled the night before.
Where AI investments fail to deliver
None of this works automatically, and the honest case for AI in brokerage has to account for how often it doesn't pay off. Most brokerages deploying AI today are automating the work a client can see rather than the work that actually creates value, and that misplaced priority is why so many pilots produce an impressive demo and no measurable change in deal outcomes. MIT research found that a large majority of generative AI pilots across industries delivered no measurable impact on profit and loss despite real investment, and the pattern behind that failure is specific: firms reached for sales and marketing automation first, while the largest documented returns actually came from back-office work and data extraction. Brokerage AI adoption has followed the same pattern: it chases the visible, client-facing use cases before the unglamorous data work that underpins them.
Two structural problems compound that mistake. The 2026 DNA of CRE report found that brokers typically manage prospecting, marketing, and transactions across roughly seven separate software tools, and keeping information consistent across all seven is now one of the biggest operational headaches brokerage leaders report. Layered on top of that fragmentation is a data quality problem: firms that feed AI platforms clean, standardized property and financial data get materially more value out of those platforms than firms feeding them inconsistent, siloed records, and that performance gap widens the longer it goes unaddressed.
The sharpest version of the objection, and the one that matters most for a broker weighing where to start, is that market analysis an investment committee doesn't trust is worse than having no analysis at all, because now someone has to spend time rebutting it instead of simply doing the work manually. The fix is sequencing. Start AI deployment where the output is easy to check and verify, such as document drafting and data extraction, and build confidence in what the tool produces before you trust it with decisions that carry real financial stakes. AI deployments fail when brokerages skip that sequence; the underlying technology is not the problem.
The deployment sequence that gives brokers a repeatable pitch advantage
Brokers who want AI to become a lasting listing-pitch advantage build their deployment outward from the most verifiable, lowest-risk workflows, so they can trust what the tool produces before they put it in front of a client, where a mistake costs credibility. That sequence has a logical order, and it starts where this piece started: listing content. The output there is quick to review, easy to check against what a human would have written, and it gives a broker or team an immediate, concrete sense of what good AI output actually looks like.
From there, the natural next step is OM and deal documentation. Once a team trusts the drafting workflow on lower-stakes content, OM production becomes the highest-leverage place to apply it, because it compresses the single most time-consuming document in a listing pitch, and the client reads and judges it directly. After that comes financial data extraction and spreading: automating the pull of rent rolls, T-12s, and financial schedules removes the manual bottleneck that has always slowed down the underwriting narrative, and this is the layer where purpose-built CRE platforms, including Lev's workflow-specific agents for financial spreading and capital markets, separate themselves from general-purpose chat tools that need heavy configuration to handle messy, real-world documents.
With the financial narrative grounded in real, verified numbers, a broker can finally build the capital markets layer on top: which buyers are likely to transact, at what capital structure, and on what timeline. A broker who consolidates drafting, data extraction, financial spreading, and capital markets intelligence into one connected platform, instead of stitching together seven disconnected tools, compounds the advantage at each step, because the output of one workflow becomes the input to the next without anyone re-entering data by hand.
Brokerages that have already built out this kind of sequenced, workflow-first deployment are reporting results that go beyond the demo stage: listing content produced at scale, faster response times to new leads, and measurable reductions in cost per closed side. The sequence matters because each stage builds the trust the next stage depends on, and skipping ahead to the client-facing work without that foundation is how a pilot turns into one of MIT's unmeasured, unprofitable deployments instead of a repeatable pitch advantage.
