The AI conversation inside real estate has two modes, and most of the professionals reading this article have encountered both of them. The first mode is vendor enthusiasm. A CRM company, a lead generation platform, or a marketing vendor has told you that artificial intelligence will revolutionize your practice, and the product they are selling is the revolution. The second mode is industry anxiety. A peer, an article, or a broker meeting has told you that AI will replace real estate professionals within a few years and that you need to adapt immediately or lose your career.
Both modes miss the reality that has been developing across the industry since the public release of ChatGPT by OpenAI in November 2022, the broader rollout of generative AI tools across 2023 and 2024, and the integration of those tools into platforms including Microsoft Copilot, Google Gemini, Anthropic’s Claude, and the vertical-specific products that have emerged across the real estate technology sector. The reality is more specific, less dramatic, and more operationally useful than either mode suggests. Certain tasks are being automated quickly. Certain tasks are resisting automation in ways that may surprise professionals who assumed those tasks were the most vulnerable. The positioning question for a working professional is not whether AI will affect her practice. The positioning question is which specific tasks will shift, which specific tasks will become more valuable, and what she should be doing right now to adjust her practice for the landscape that is developing.
The tasks AI is already automating well
Several categories of real estate work have been meaningfully automated by tools that are available today, and the automation is improving with each model release. Understanding the specific categories where the automation has moved fastest is the starting point for any honest assessment.
The first category is the generation of written content across a practice. Listing descriptions, property marketing copy, email follow-ups, newsletter content, social media captions, and initial drafts of client communications have all been automated effectively by general-purpose large language models and by vertical products built on top of them. A professional who uses ChatGPT, Claude, or Gemini to draft a listing description produces a usable first draft in under a minute, compared to the fifteen to thirty minutes a human-only process would typically take. The quality of the output has reached a level where the human edit is substantive, meaning adjustments for voice, specific detail, and local context, but no longer requires starting from a blank page. The productivity gain is measurable and is available to any professional willing to integrate the tools into her workflow. Products including Luxury Presence AI features, BoomTown content tools, and many CRM systems have integrated generative writing capabilities directly into the professional workflow, making the tools accessible without requiring the professional to maintain a separate AI subscription.
The second category is comparative market analysis and preliminary pricing work. AI-assisted automated valuation models have existed for years through products including Zillow’s Zestimate, Redfin Estimate, and Realtor.com’s pricing tools, and the integration of more sophisticated machine learning approaches into commercial-grade automated valuation models through platforms including CoreLogic and HouseCanary has improved the accuracy of preliminary valuation significantly. The research on automated valuation model accuracy, including studies published by the Federal Housing Finance Agency and academic research in journals including the Journal of Real Estate Research, has documented that commercial-grade automated valuation models produce median error rates within a narrow single-digit percentage range on typical residential properties in data-rich markets, though error rates increase in unusual property types and in markets with less transactional data. The practical effect for a professional is that preliminary pricing work, which previously required one to two hours of comparable property research, can now be generated in minutes as a starting point. The professional’s adjustment work on top of the automated valuation, incorporating the specific characteristics of the property, the specific condition factors not visible in the data, and the specific market dynamics at the micro level, remains essential to a credible listing price, but the starting point is delivered by the technology.
The third category is administrative task automation across the transaction. Scheduling, appointment setting, follow-up reminders, document organization, transaction coordination task tracking, and client communication cadence management have been automated through products including Follow Up Boss workflow automations, Sierra Interactive’s automation tools, and newer AI-powered transaction coordination products. The research on where professional time is actually spent, including survey data published by the National Association of Realtors across its member activity reports, has consistently documented that administrative work consumes a substantial portion of a real estate professional’s working week. Moving even a meaningful portion of that administrative burden to automated systems frees the professional for work that automation does not do as well.
The fourth category is marketing personalization and lead nurture at scale. Products including zBuyer, Ylopo, and the AI-assisted features in platforms like kvCORE have integrated machine learning models that personalize content delivery to leads, predict which leads are most likely to convert, and automate the initial nurture sequences that previously required manual attention from a professional or an assistant. The quality of the automation in this category varies significantly across products, and the evaluation of specific tools requires hands-on testing in a professional’s own practice, which is consistent with the broader pattern of evaluating any marketing technology product.
The tasks that are resisting automation, and why
A widespread assumption in the AI conversation is that the tasks most easily automated are the ones that look simple on the surface. The pattern has actually been the reverse in many cases. Certain tasks that appear simple have resisted automation, while certain tasks that appear sophisticated have been automated quickly.
The tasks most resistant to automation across real estate share a specific pattern. They require judgment under conditions of uncertainty, they depend on relationship context that is difficult for a model to access, and they involve the management of human emotion during high-stakes decision-making. Consider the tasks a real estate professional actually performs in the critical hours of a difficult negotiation. The reading of micro-expressions and tone across a conversation. The calibration of what to say and what to hold back based on a seller’s emotional state. The navigation of a low appraisal with parties whose financial pressures are pulling in opposite directions. The judgment about when to push a client toward a decision and when to give her space. None of these tasks has been meaningfully automated by current AI capabilities, and the research on what current models can and cannot do, including work published by researchers at Stanford University’s Human-Centered AI Institute and at the Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Laboratory, indicates that the gap between language generation and high-stakes relational judgment remains substantial.
Erik Brynjolfsson, professor at Stanford University and director of the Stanford Digital Economy Lab, and his longtime collaborator Andrew McAfee at the Massachusetts Institute of Technology, have published extensively on the pattern of what automation absorbs and what it does not. Their 2014 book “The Second Machine Age” and their subsequent research, including Brynjolfsson’s ongoing work through the Stanford Digital Economy Lab, have consistently documented that automation tends to absorb tasks with well-defined inputs, well-defined outputs, and large training datasets, while leaving tasks that require judgment, context integration, and human trust more intact. The finding applies directly to real estate. Listing description writing has well-defined inputs, well-defined outputs, and large training datasets. A closing-table conversation with a seller whose spouse has just received difficult medical news does not.
Daniel Susskind, economist at King’s College London and author of the 2020 book “A World Without Work,” has argued across his work that the professions most affected by automation tend to lose their routine, codifiable tasks first, while their judgment and relational tasks become proportionally more central to the professional’s value. The framing is useful for real estate professionals trying to understand how to position for the next three to five years. The tasks that automation absorbs become, in effect, table stakes. Every professional will have access to the same automated listing description tools, the same automated valuation models, and the same administrative automation. The professional’s differentiated value will increasingly rest on what she does that the automation does not do, and the list of those activities is both longer and more specific than the industry conversation has typically acknowledged.
What to position for across the next three to five years
The operational positioning for a real estate professional follows from the analysis above. Three adjustments are worth making in the current year.
The first adjustment is active competence in the AI tools that are automating the table-stakes tasks. A professional who is not using generative AI for written content in 2026 is operating at a productivity disadvantage against professionals who are. The learning curve is short. Spending two hours a week for a month with a general-purpose tool like ChatGPT, Claude, or Gemini and a specific list of use cases drawn from the professional’s own practice produces functional competence that compounds through every subsequent week of use. The professional is not adopting AI to replace her judgment. She is adopting AI to move the routine tasks off her calendar so that her judgment has more time to engage with the tasks where judgment actually matters.
The second adjustment is a deliberate investment in the skills that become more valuable as automation increases. Negotiation skill, relational depth with clients, expertise in complex transactions, the ability to explain financial and regulatory nuance in plain language, and the credibility that comes from years of accumulated judgment in a specific market all become more valuable, not less, as routine tasks are automated. David Autor, professor of economics at the Massachusetts Institute of Technology, has published extensively on how technological change affects labor markets, and his research, including studies published in the American Economic Review and summarized in his 2015 paper “Why Are There Still So Many Jobs? The History and Future of Workplace Automation” in the Journal of Economic Perspectives, has consistently shown that professionals who invest in complementary skills to emerging technologies capture more of the economic value created by those technologies than professionals who either ignore the technology or compete directly with it. For a real estate professional, the complementary skills are clear. Automation handles the routine. The professional handles the judgment, the relationships, the context, and the trust.
The third adjustment is skepticism toward vendor claims about AI capability. Many of the products being marketed with AI branding across the real estate technology sector are incremental improvements on existing products rather than transformative new capabilities, and the responsible evaluation of any AI product requires the same disciplined, hands-on testing that applies to any technology purchase. The short list of relevant evaluation questions is consistent across categories. What does the tool actually produce in practice, measured against a specific workflow. What does it cost in time and money relative to the time and money it saves. Where does it break or produce low-quality output that requires manual correction. Who is responsible for the data the tool ingests and the outputs it generates, particularly in regulated contexts including fair housing compliance, which the Department of Housing and Urban Development has indicated it is actively monitoring in the context of AI-powered marketing and lead generation tools. A professional who asks these questions before adopting a tool avoids the most expensive version of the AI mistake, which is paying for a product that does not deliver what the marketing suggested.
The AI conversation in real estate will continue to evolve rapidly across the next three to five years, and professionals who stay current through direct, hands-on testing of new tools and through attention to credible research on the state of the technology will be better positioned than professionals who rely on secondhand opinions. The tasks AI is automating well are clearly identifiable today, and the tasks AI is not automating well are equally identifiable. The professional’s work is to move the first category into automated workflow, to invest deliberately in the skills that matter more as the automation spreads, and to evaluate new claims with the same skepticism she would bring to any technology purchase. The professionals who do all three will be the ones whose practices strengthen through the transition. The professionals who do none of the three will be the ones asking, in 2029, what happened to the market position they once held, and finding the answer in the specific adjustments they declined to make in the mid-2020s.
