AI adoption in real estate is now nearly universal. 92% of real estate firms are piloting AI, and 97% of institutional investors report it integrated into their investment process (JLL, 2025/2026; Dealpath, 2026). Yet only 5% of firms say they’ve achieved most of their program goals, and the share of executives calling AI’s impact “transformative” has fallen from roughly 12% to about 1% year over year (Deloitte, 2026).
Quick facts
For the fuller landscape of how AI fits into commercial real estate operations more broadly, see our guide to AI in commercial real estate.
Right now, AI in real estate is mostly doing the easiest part of a broker’s day, with the part that eats it staying manual.
AI shows up most in writing copy (59%), prospect outreach email (56%), and social media content (52%). What actually takes up most of a broker’s day is a different set of activities, primarily finding new prospects (57%), client reporting (45%), data entry across systems (37%), and deal and pipeline tracking (24%), per the 2025/2026 research from JLL, Dealpath, and Buildout and theBrokerList.
Line up the two lists and the mismatch is hard to miss. The tasks getting AI’s attention aren’t the ones eating a broker’s calendar. Call it the wrong-hours problem. AI has automated the part of the day that was never the bottleneck, while the part that costs brokers their time stays manual.
“ AI gives back the wrong hours. It makes producing a document cheap while leaving the expensive work untouched: finding the right source material, deciding what is actually true, verifying every claim before it travels. The minutes you save at the keyboard come back as hours spent checking the output. ”
92% of CRE companies are piloting AI, but only 5% report achieving most of their program goals (JLL, 2025/2026), and the share of executives calling AI’s impact “transformative” fell from roughly 12% to about 1% year over year (Deloitte, 2026, N=850). Two things explain most of that gap.
First is the wrong-hours problem from above. A company can roll AI out everywhere it currently fits and still not see it move the numbers that matter, because those numbers come from prospecting, reporting, and deal tracking, the work AI barely touches yet.
Second, “92% piloting” hides how shallow adoption still is for a large share of the market. 17% of brokers report using no AI at all, and 53% run deals with no dedicated management system whatsoever (Buildout and theBrokerList, Q4 2025/2026). Without a system built around clean deal data, one where AI and daily workflow connect, AI stays bolted onto the outside of the job. Employees are left re-explaining context and copy-pasting output between disconnected tools every time. That does not just eat up additional time, it demands additional effort, making AI a harder habit to reach for in the first place.
One caveat is worth adding. Some of this gap is simply early. Deloitte’s 2026 outlook also found that AI returns often take longer to show up than expected, with how fast people adapt, not the technology itself, acting as the limiting factor. Part of the “transformative” drop is likely that catching up, not only failed pilots.
A manual check is still required before the output can be used, and for many teams that check likely cancels out whatever time AI saved to begin with. 41% of companies call AI-assisted work a net time cost once verification is factored in (JLL and Dealpath, 2026).
Verification here is not optional. Real estate runs on sensitive, high-stakes data, financial terms, ownership records, tenant information, and no business can put an AI-generated number or clause into a contract or a client conversation without a human confirming it first. Liability for an error sits with the company, not the model, which is likely part of why regulatory and compliance concerns show up as a top adoption barrier in Dealpath’s own research.
Where the tax is lighter, the payoff shows up fast. CBRE expects to cut research costs by roughly 25% using AI (Bisnow, 2026), and research is exactly the kind of task where that trade works. Summarizing existing information is quicker to check than a valuation or lease term a client will actually act on, which keeps the stakes of an occasional error much lower. Verification pays off fastest wherever it is cheapest to do.
CRE professionals trust AI with some tasks far more than others. Document analysis is the single most common AI use case in commercial real estate, at 67% of companies, per Dealpath data cited in NAIOP and CREDA Global’s 2026 “Structure Before Speed” report, while deal valuation remains something professionals are far less willing to delegate. The pattern lines up with the verification tax above: cheap to verify pays off quickly, expensive to verify stalls.
There’s a reasonable explanation for that split. Reading a document is a bounded task with a checkable answer, either the AI found the right clause or it didn’t. Valuing a deal isn’t bounded the same way, it folds in judgment, local market knowledge, and relationships that don’t reduce cleanly to a number a model can output with confidence. That’s less a shortcoming of the technology and more a mismatch between what today’s tools are built to do well and what the highest-stakes real estate decisions actually require. Whether that gap narrows as models and companies mature, or whether valuation stays a human call by design, is one of the more interesting open questions in this space over the next few years.
Underneath both the verification tax and the trust gap sits an infrastructure problem. 81% of CRE companies report three or more AI systems that aren’t producing expected results, 43% point to fragmented data as the top reason AI underdelivers, and roughly 90% of failed deployments trace back to infrastructure rather than the model itself (JLL and Dealpath, 2025/2026). 53% of brokers, as noted above, are working without even a basic deal-management system.
Fragmented data is as much a usability problem as a storage one. Deloitte’s 2026 outlook found that having more data doesn’t guarantee AI can use it, the real bottleneck is often finding significant, usable data without extensive extract-transform-load work. Real estate data frequently includes sensitive details, bank account and social security numbers, tenant names, mortgage payment statuses, that can’t be fed directly into a model for training.
Nearly half the companies Deloitte surveyed named synthetic data generation as a high-interest fix for exactly that reason, though building it well takes data science expertise most CRE teams don’t have in house, per the same Deloitte research. The upshot: businesses sitting on more data than any AI vendor could want are still, in a meaningful sense, data-poor, because volume was never the constraint that mattered.
Companies getting better ROI from AI built the foundation first. The ones with a mature technology program already in place, meaning established data infrastructure, change management, and experienced teams, get considerably more from their AI efforts than businesses without that foundation (JLL, 2025/2026). Organizations that redesign workflows before selecting an AI tool are twice as likely to see significant financial returns, a McKinsey finding cited in NAIOP and CREDA Global’s 2026 “Structure Before Speed” report.
Most companies are starting from behind on exactly that foundation. Over 60% report having to fix basic technology issues, duplicated functionality, dormant systems, before they can use AI fully at all (JLL). That’s a double burden, catching up on fundamentals while trying to compete on AI at the same time.
JLL’s survey points to a practical way around that. Respondents named the best time to adopt or change a technology system as alongside a major change already underway, an IT overhaul, a leadership change, a new regulatory requirement, or a capital planning cycle. Rolling AI out alongside a change that’s happening anyway is one of the more consistent ways companies report securing the resources and buy-in a new system needs.
Adopting AI in real estate comes down to three things: enough trained people, budget that survives scrutiny, and governance over how the tools actually get used day to day.
Training is the most visible gap. Only 33% of the CRE workforce feels adequately trained on AI, and 43% name a lack of internal AI expertise as the single biggest adoption barrier, ahead of both cost and regulation (JLL; NAIOP and CREDA Global, 2026). Dealpath’s State of AI Readiness research lists the same barrier near the top, alongside regulatory and compliance concerns, budget constraints, and decentralized data.
Budget sits right behind it. 65% of organizations report CRE tech budget pressure over the past two years (JLL), driven by both the talent gap above and by ROI expectations that now demand a fuller business case before any AI investment gets approved.
There’s a less obvious layer underneath both, governance over how AI is used day to day. Token and inference cost unpredictability has been flagged as a live concern by JLL’s own Global Chief Data Officer (Bisnow, 2026), and that is a training problem as much as a technical one. A team that doesn’t know which model suits which task, or how to prompt efficiently, tends to reach for more computing power than a job actually needs, without getting more value out of it.
The same logic is pushing companies toward smaller, purpose-built models instead of one large general-purpose system for everything, a shift we return to later in this piece. A narrower model is easier to budget for and easier to govern than a general one, which is exactly the kind of control most CRE teams are still building toward.
Not all non-adoption is passive: part of it is a deliberate call, not simply lagging behind. 17% of brokers report using no AI at all (Buildout and theBrokerList, Q4 2025/2026). Some of that is inertia, but not all of it is passive. Real estate data is often sensitive by nature, the same bank account numbers, social security numbers, and tenant records that make internal data hard to use for AI training also give companies a reason to be wary of exposing it to a third-party model at all (Bisnow, 2026). A portion of the non-adoption number likely reflects a deliberate call about what a business is willing to hand a vendor, not simply lagging behind.
CRE isn’t alone in this. A Gartner analysis of AI-ready data found that 60% of AI projects built on ungoverned data are at risk of abandonment (Gartner, Feb 2025), and S&P Global tracked the share of companies walking away from most of their AI initiatives climb from 17% to 42% in a single year (S&P Global Market Intelligence, 2025, via Cybersecurity Dive). The adoption-and-results gap in real estate looks less like an industry-specific failure and more like a pattern the wider corporate world is living through too.
If you’re past asking whether this is happening and want to know what fixes it, our related piece on real estate AI adoption challenges covers that next.
Once buyers move past piloting, industry expertise ranks as the single biggest factor in vendor selection, at 52%, ahead of both price (49%) and product fit (46%) (Responsive, 2026). That finding comes from B2B buyers broadly rather than real estate specifically, so it’s worth reading as a signal rather than a confirmed preference. Real estate has its own reason to lean the same way: 81% of CRE companies have already been burned by an underperforming system, and a vendor’s proof of real-estate-specific implementation is likely to carry more weight than a generic AI pitch as a result.
The clearest AI adoption trend in real estate for the next 12 to 18 months is a narrower approach than the pilot-everything phase most companies are leaving behind. Deloitte’s 2026 outlook found CRE organizations shifting from broad AI strategies toward targeted deployments in tenant relationship management, lease drafting, and portfolio management specifically.
That shift also comes with a move away from one large, general-purpose model, toward industry-specific platforms such as IWMS or CAFM systems (about 22% of companies), and public LLMs fine-tuned or paired into smaller, task-specific systems (another 20%). Narrower, in other words, is starting to beat bigger.
Across every section here, the same shape keeps repeating. Adoption is already close to universal, but most of it is aimed at the easiest, lowest-stakes part of the job. Verification is where AI earns or loses its keep, and the companies pulling ahead are the ones that fixed data, workflow, and training before chasing new tools. The next 12 to 18 months look less like a race to adopt more AI and more like a shift toward using it more deliberately.
That kind of deliberateness is also the lens behind Ascendix’s own AI consulting and technology-partner work, which starts from real estate operations rather than a generic AI rollout.
Have a Question About AI in Your Operations?
Our team answers these directly, real estate is all we do.
Broker AI usage rose from 38% to 83% between 2024 and 2026, though 17% still report using none (Buildout and theBrokerList, Q4 2025/2026). Across CRE companies broadly, 92% are piloting AI in some form (JLL, 2025/2026). See our related piece on how real estate agents are using AI day to day.
For many teams, it’s closer to shifting the work. 41% call AI-assisted work a net time cost once verification is factored in, and only 51% say AI saves time once that check is included (JLL and Dealpath, 2026).
Mostly infrastructure, not the model itself. 81% of CRE companies report three or more AI systems not producing expected results, and roughly 90% of failed deployments trace back to fragmented or unusable data rather than the tool (JLL and Dealpath, 2025/2026).
Traceability. AI is trusted for document analysis, the single most common CRE use case at 67% of companies, but far less trusted for high-stakes calls like deal valuation, where the cost of an unverified error is too high to skip a manual check.
Industry expertise ranks as the top vendor-selection factor among B2B buyers broadly, ahead of both price and product fit (Responsive, 2026). Real estate buyers have their own reason to weigh it just as heavily: 81% of CRE companies have already been burned by a system that didn’t deliver, so proof of real-estate-specific implementation tends to matter more than a generic pitch.
We are a team of CRM consultants, developers, data analysts from the United States and Europe. Since 1996, we've been helping companies make the most out of CRM software and improve their software systems.
(No Ratings Yet)
Loading...