Enterprise AI trends for CRE in 2026: what the surveys actually show

Enterprise AI in commercial real estate entered 2026 with the adoption argument settled and the results argument barely started. Nearly every CRE team is piloting something. Almost none report hitting the goals they set. The eight trends below each rest on a named figure from one of two industry surveys rather than on a prediction.

Two sources carry this piece: JLL’s Global Real Estate Technology Survey, published 27 October 2025, drew on more than 1,000 senior CRE decision-makers across 16 markets. Deloitte’s 2026 Commercial Real Estate Outlook surveyed over 850 C-level executives at owners and investment companies holding at least $250 million in assets under management, fielded in June and July 2025.

Adoption is settled. Outcomes are not.#

92 percent of CRE teams have started piloting AI or plan to this year, up from under 5 percent three years earlier. In the same JLL survey, 5 percent report having achieved most of their program goals. An 87-point gap between starting and succeeding is explained by what happens after a pilot produces a good answer. Before the next pilot begins, write down the completed action it has to produce, not "evaluate AI" but a specific finished output someone uses without rework.

Deployment is narrowing to named workflows#

Deloitte found the top three AI focus areas for the next 12 to 18 months are tenant relationship management, lease drafting, and portfolio management, with organizations shifting toward targeted deployments. Lease drafting only pays if the negotiated term reaches the abstract, and the abstract reaches the charge schedule, which is the substance of lease abstraction.

General-purpose tools are doing real estate work#

20 percent of Deloitte’s respondents use publicly available large language models, against 22 percent using an industry-specific platform. Those two numbers being close is the story: a fifth of large owners are putting real estate questions to general tools with no access to lease terms, charge history, drawings, or operating records.

Data sensitivity is shaping architecture#

Roughly half of Deloitte’s respondents identified synthetic data generation as an area of significant interest, driven by concerns about using sensitive real estate data in AI training. Ask any vendor where your documents go, whether they are used for training, and what separation exists between your data and anyone else’s. Our own position is documented at trust.

The legacy stack is the binding constraint#

81 percent of companies report at least three existing systems that are not generating expected results, and 88 percent are allocating budget to upgrade legacy technology. A layer that only produces more answers becomes the fourth underperforming system, which is why working over an existing system of record is only sound if the addition finishes something.

Budgets tightened while adoption rose#

65 percent of organizations report CRE technology budget pressure over the past two years, and more than half report longer procurement decision-making periods. That combination produces small pilots that are easy to approve and hard to scale. Build the business case around a countable result rather than estimated hours saved.

Training is the unaddressed bottleneck#

Only 33 percent of the workforce feels adequately trained on AI. Judge tools by what they do without being asked. If the output arrives inside the work someone is already doing, training matters less.

Sourcing is mixed, not single-vendor#

70 percent of occupiers use multiple sourcing strategies for AI capability. Ask how each candidate behaves as one component among several: what it reads, what it writes back, and what happens to your data if you stop using it. Portfolio-level decisions tend to draw on more sources than any single system holds, which is the shape of portfolio strategy work.