The State of AI Adoption in Commercial Real Estate Portfolios
The numbers on AI adoption in commercial real estate are remarkable — and, on close inspection, somewhat contradictory.
An impressive 88% of real estate investors, owners, and landlords are already piloting AI, according to JLL's 2025 Global Real Estate Technology Survey — a figure that has risen sharply from under 5% of CRE teams in 2023. However, just 5% of respondents said they have achieved all their AI program goals.
What to make of this disparity? It seems clear that although many organizations are now doing something with AI, precious few have much to show for it yet.
This gap — between the pace of adoption and the pace of results — is among the most interesting stories in real estate technology right now. Understanding it requires looking past headline adoption rates and asking more precise questions: Where is AI actually being deployed? What's blocking it from working? And what separates organizations that are starting to see real returns? We’ll explore these and more in the article below.
The Adoption Surge Is Real — But Shallow
The jump from single-digit to near-universal piloting in less than two years is striking by any industry standard. JLL has identified 28 AI use cases across the real estate value chain, with most companies actively pursuing five pilot projects simultaneously. If anything, it’s evident that the activity level with respect to AI adoption is high.
But activity and maturity are different things. While adoption is widespread, maturity remains low. The real challenge lies in the readiness of an organization, including whether they have quality data, infrastructure, and the change-management processes needed to integrate AI into core workflows.
McKinsey's recent analysis of agentic AI in real estate puts a finer point on the problem: AI adoption is widespread across industries, yet scaled bottom-line impact is hard to find, often because tools sit adjacent to workflows instead of being embedded within them. That description fits the CRE sector precisely. Most organizations have launched sensible pilots — summarizing a lease, drafting a memo, answering a question faster — without changing how work actually moves through core systems.
In facilities management specifically, 28% of organizations have embedded AI solutions in their FM operations — rising to 46% for large organizations with 100,000 or more employees — with many advancing from initial pilots to scale across multiple function areas and facilities.The gap between large and mid-sized portfolios is notable. Scale appears to confer an advantage, both in resources to invest and in the volume of operational data that makes AI systems more useful.
Energy Management Is Where Operational AI Is Landing First
For building operators and engineers, the use case with the most traction is energy management — and for understandable reasons. Energy is one of the largest controllable operating expense lines in a commercial portfolio, and the potential savings are material. AI can cut energy and maintenance costs 10% to 30% through predictive controls, according to JLL's occupier trend analysis, though ranges this wide reflect significant variation in building type, baseline conditions, and implementation quality.
The academic literature offers more granular benchmarks. A 2025 meta-analysis published in Energy Informatics, reviewing 126 peer-reviewed studies spanning 2010–2024, found that AI-driven optimization achieves substantial energy savings, with hybrid methods showing the highest potential (28.1% ± 12.3%), reinforcement learning providing strong consistent performance (22.3% ± 8.4%), and supervised learning offering reliable but more modest gains (14.7% ± 5.2%). These figures come from controlled studies, so real-world portfolios should apply them with judgment, but the directional evidence is consistent.
The same review flagged a structural constraint that anyone working in building operations will recognize: insufficient multi-system integration, with 76% of studies examining single-system implementations. Most AI energy optimization work, in both research and practice, still operates on one system at a time — one AHU, one chiller plant. The harder and more valuable problem is coordinating optimization across systems within a building, and across buildings within a portfolio.
Data Quality Is the Bottleneck Nobody Likes to Talk About
The gap between AI adoption and AI outcomes can largely be traced to a single underlying problem: data. Not a lack of data — commercial buildings generate substantial operational data — but data that is inconsistent, poorly labeled, trapped in siloed BMS environments, or simply not structured in ways that AI systems can use.
Legacy systems represent key barriers to AI adoption, with 81% of companies reporting at least three existing systems that aren't generating expected results, and 88% allocating budget to upgrade legacy technologies.
54% of CRE teams cite compatibility issues with legacy infrastructure as the top barrier to AI progress. This shows up concretely in building operations: inconsistent point naming across BMS vendors, gaps in sub-metering, and the absence of a standardized ontology that would let an AI system understand that "CHW-SPS" and "chilled water supply" refer to the same thing in different buildings. Before any meaningful optimization can happen, that foundation has to be in place.
McKinsey frames this as a prerequisite for the entire domain transformation agenda. A "factual layer" — which makes real estate data and documents usable by collecting clean property, system, vendor, and lease metadata, and serving as a clear source of truth when systems disagree — is the first of five technical layers required for agentic AI to succeed. Without it, even well-designed AI systems can't operate reliably.
Who's Getting Results (and Why)
Rather than enabling laggards to leapfrog ahead, AI adoption is widening the already-present gap between technology leaders and everyone else. Companies that already run successful tech programs are pulling further ahead in AI outcomes.
This may seem like a counterintuitive finding, but it makes sense upon further consideration: more tech-forward organizations have already invested in building data infrastructure, operational discipline around technology, and internal literacy around analytics; it follows logically that they are now the ones primed to get the most out of AI applications.
WIth that in mind, the most successful teams are those taking a structured and consistent approach: developing tailored roadmaps, investing in high-quality data, and embedding AI into everyday workflows. That applies equally to financially-minded portfolio-level roles and building operations teams. The organizations making progress share a common pattern: AI is thoughtfully integrated into recurring operational processes, not treated as a standalone project.
McKinsey's framework for where this integration works best offers a useful organizing principle. Rather than asking "what use cases can we pilot?", successful real estate leaders getting results are asking "which workflows should we redesign so the software is allowed to do the work, with appropriate controls?" The distinction matters. A use case is bounded and narrow, while a workflow redesign changes how work actually moves.
The same framework distinguishes between two types of tasks within any operational domain: "steps" — repeatable tasks that benefit from speed, consistency, and clean handoffs — and "thoughts," which are judgment calls that require discretion, such as exceptions, trade-offs, and decisions that carry financial, reputational, or regulatory risk. The practical implication: automate steps aggressively, and protect the moments that require human judgment. This is a more useful frame than "AI vs. human" for thinking about where technology actually fits in building operations.
Continuous Commissioning as an AI Use Case
One area where this model is well-developed is continuous commissioning (CCx). Unlike a one-time recommissioning engagement, CCx relies on persistent monitoring of building system performance — tracking the drift between design intent and actual behavior across AHUs, VAV boxes, chilled water systems, and controls — and triggering action when performance degrades.
Through the steps-and-thoughts lens, CCx is almost entirely a steps problem: pattern recognition at high volume, continuously, across data streams that human teams cannot monitor at the same frequency or scale. The operational savings come not from a single intervention but rather from the elimination of persistent faults and inefficiencies that would otherwise accumulate undetected between manual inspections. Human engineers retain the judgment calls, deciding which faults to prioritize, weighing comfort against efficiency targets, and escalating anomalies that fall outside expected parameters, though the analysis of key data inputs that makes such prioritization faster and easier (e.g. assigning a financial impact to a given fault; grouping faults with others that have the same root cause into a single, larger bucket of work), can be automated, too.
For portfolios that already have the data infrastructure in place, CCx represents a relatively high-confidence AI application: the use case is well-defined, the performance signal is clear, and the comparison baseline (metered energy against weather-normalized expectation) is auditable.
The Agentic Layer Is Already Taking Shape
The current wave of AI deployments is not the final form. Agentic AI is accelerating beyond previous applications by leveraging true reasoning models to automate multistep workflows inside core business systems, enabling humans to work in partnership with AI agents. The shift is from "help me understand" to "help me get it done.”
In building operations, the implications are concrete. Rather than tools that stop at flagging a potential chiller fault for an engineer to investigate, an agentic AI system can identify the fault, validate its root cause, find related issues, group them all into a project (and autonomously draft the details and recommended resolution steps), and route it to the right person – enabling the engineer to simply review, approve, and act. Unsurprisingly, organizations that have automated such processes have seen time savings of more than 30% on many workflows.
A labor productivity analysis by the McKinsey Global Institute suggests that automation, including AI applied to knowledge work, could unlock roughly $430 billion to $550 billion in annual value globally across real estate, construction, and development. That figure encompasses the full value chain — but the building operations layer, where the volume of daily decisions is highest and the data infrastructure is most mature, is a natural place for early gains to accumulate.
The Realistic Near-Term Picture
The dominant story in CRE AI adoption right now is experimentation, with meaningful results concentrated in organizations that had already done the foundational work. According to JLL, more than 60% of real estate investors remain unprepared strategically, organizationally, and technically for true AI integration.
That is not an argument for waiting. The organizations building advantage now are doing so precisely because they started earlier and moved deliberately. The purpose of current pilots isn't just immediate ROI, but also providing critical learnings to inform more encompassing, longer-term strategies.
For building operators and portfolio managers, the near-term priorities are consistent across the research: invest in data infrastructure before expanding AI tooling, focus AI on well-defined operational problems rather than vague transformation mandates, and treat energy management and facilities workflows as the use cases most likely to generate near-term, measurable returns. The transition to agentic AI will raise the stakes for that foundational work; organizations that have clean, connected, normalized building data will be positioned to move quickly, while those that don't will find the gap harder to close.
When the dust settles, the winners will be the teams who prepared early and deployed technology strategically, and the providers whose platforms quietly get the work moving before the day begins, so human operators can focus on judgment, relationships, and the moments that matter in the built environment – as McKinsey puts it, automating steps and protecting thoughts.
About Noda
Noda is a data and analytics company on a mission to make every building smarter, more efficient, and more sustainable. Recently ranked in the top 10 tech companies leading the charge on climate action, its AI-powered suite of products surface unique insights that empower real estate teams to reduce costs, decrease time spent on routine work, and find and act on opportunities to save energy and carbon. Discover how Noda's solutions can unlock the potential of your assets and accelerate the transition to net zero. Visit us at noda.ai to learn more.