Every year, Realcomm is a useful gut check on where the industry actually is versus where the coverage says it should be. This year, the gap between the two was smaller than usual. Across multiple sessions, a consistent argument emerged that had less to do with any single technology and more to do with a structural shift in how real estate firms are thinking about AI. The conversation has moved. It is no longer about whether to adopt AI. It is about what kind of AI adoption actually produces results, and what the firms that are getting results have figured out that others have not. Increasingly, firm size has less to do with that answer than you might expect.
The pilot era is over. The architecture question is what comes next.
The session on AI readiness, framed around four pillars of organizational preparedness, surfaced something that most practitioners in the room already knew but rarely say out loud: pilot hell is real, and the reason most pilots never reach production has almost nothing to do with the AI. It has to do with the infrastructure underneath it. Data governance, cloud connectivity, risk appetite, talent availability. These were the friction points being named, and they were being named not as prerequisites to delay deployment but as the actual work that separates firms running AI in production from firms perpetually preparing to run AI. That preparation trap is not evenly distributed. The firms most likely to get stuck in it are the ones with the most complexity to navigate. For a closer look at what that gap means for mid-market firms specifically, check out this post by our team: Don’t Go Changing: The Mid-Market’s Golden Opportunity.
One panelist put it plainly: business owners have to understand that they are the true information owners. IT owns the infrastructure and the processes, but the data belongs to the people running the business, and until they accept that accountability, data governance stays stuck. Another observation that landed hard was this one: most executives believe AI deployment is easier than it is, because the people who built the thing rarely explain what it took. Closing that gap, setting realistic expectations without losing momentum, is one of the defining leadership challenges of the current moment.
Task automation and workflow automation are not the same thing.
The session on moving AI from experimentation to daily operations drew a line that more conversations in the industry need to draw more clearly: the difference between augmentation and automation, and the difference between making one step faster and changing the workflow underneath it. The examples in the room ranged from voice agents handling incoming leads and maintenance requests in multifamily to structured prompt sequences generating analyst-ready report summaries. Each of them worked. None of them, on their own, was the point.
The point, as one speaker put it, is that instructions and codified business processes are the two things that build real automation. If you can encode how your organization makes a decision, you can automate around it. If you cannot, you have a faster task inside a broken workflow, and the bottleneck has just moved one step downstream. The firms seeing the most traction were the ones who started with that encoding work first, and then built the AI layer on top of it. Not the other way around.
The question is no longer whether legacy systems survive. It is what sits on top of them.
The panel on AI and the future of real estate technology, which brought together operators, investors, and platform executives, produced the most direct conversation of the conference on this topic. The session opened with a show of hands: how many people believe all the software in real estate will eventually be replaced by AI? A handful. How many have already replaced one legacy or point solution with an AI-native alternative? More, but still a fraction of the room.
The consensus that emerged was neither utopian nor defeatist. Systems of record, the ERPs and property management platforms that house the operational data of real estate firms, are not going away. What is changing is the experience of them, and more importantly, what sits above them. One operator who has been building AI integrations into his company’s workflow layer since 2018 described watching five or six AI applications grow to hundreds, activating 100,000 times a month across every department. His point was not about scale, it was about architecture. He noted that the firms that built a coherent data and workflow layer early now have a foundation that new AI capabilities can plug into. However, other speakers pointed out that the advancements in AI have made it possible to build the workflow layer on top of existing solutions, while simultaneously cleaning and building the data layer.
A platform executive framed the risk issue with equal clarity: the industry is moving from AI as an insight tool to AI as an operational participant. Those have different risk profiles. When AI is doing more than surfacing information, when it is initiating action, the control framework has to change. Financial controls, cyber governance, audit trails that can hold up in discovery, kill switches at multiple layers. These are not theoretical concerns. They are the governance infrastructure that enterprise AI deployment requires, and firms that are not building it now are creating liability they have not accounted for yet.
The enterprise operating layer is the category being built right now.
Insights from prominent technology and platform CEOs in the industry reinforced the structural argument from a different angle. The conversation kept returning to a specific tension: intelligence versus action. AI that suggests what to do versus AI that does it. Deterministic outputs versus probabilistic ones. Human-in-the-loop as a trust-building mechanism versus human-in-the-loop as a permanent limitation. The executives in that conversation were not treating these as binary choices. They were treating them as a maturity curve, with different firms sitting at different points on it depending on their data infrastructure, risk appetite, and organizational readiness.
Two observations from that conversation are worth holding onto. The first: proprietary business logic, your fund hierarchies, your underwriting assumptions, is becoming a competitive asset in a way it was not before. One framing that came up was the Salesforce headless CRM concept, where the data structure itself has inherent value, and firms are beginning to treat their operational data the same way. The second: the walled garden is ending. Platforms that have historically kept data locked inside their own systems are being pushed toward interoperability, and the firms that benefit most from that shift will be the ones who already have a layer capable of acting on data regardless of where it lives.
One exchange in the session on AI and legacy systems captured the broader argument well. An operator described what happens when a deal comes in: before any person touches it, a series of AI interventions runs automatically, abstracting it, checking for duplication, scoring it against investment theses, routing it to the right team. By the time an analyst sees it, they are not doing triage, they are doing judgment, and that distinction is exactly what full workflow automation is supposed to produce: people working at the level of decision-making and strategy rather than routing and validation, because the layer underneath them is built to handle the handoffs.
What the room agreed on leaving.
The lightning round takeaways from the AI and legacy systems panel said more than most conference closing remarks do. Stay focused on facts over hype. Go talk to your teams about where the manual processes actually are, because the people doing the work understand the use cases better than leadership does. Fix the data first, because AI is a mirror: if the outputs look wrong, the problem is almost always in the inputs. And perhaps most pointedly: individuals are already more comfortable with AI than companies are. The corporate ability to change is the rate-limiting factor now, not the technology.
That last one is worth sitting with. The conversation at Realcomm 2026 was not about whether real estate needs AI. That argument is over. It was about whether firms are building the right foundation to make AI do something consequential, and whether the organizational structures, the governance, the data ownership, the workflow architecture, are keeping pace with what the technology can already do. For the firms with the complexity and scale to navigate all of that, the path is long. For firms that can move without those constraints, the window to build a real operating advantage is open right now.