Every real estate organization might not be adopting AI, but they’re at least thinking about it. With all of the media noise about AI transformation and the inevitable questions from investors and advisors, it is hard for real estate organizations not to have a bit of FOMO. Some executives want to jump into action and not get stuck too long thinking about how to actually deploy it. The truth is, the “thinking about it” stage is actually very important. Assessing an organization’s AI readiness requires answering hard questions about a business’s operations, its data structure, and its strategic direction.

To start, you’ve got to know what problems you’re actually trying to solve. This is where many development teams get tripped up, according to Thomas Mackean, Digital Strategy Expert at JLL. “I am often telling our clients AI features age. Outcomes don’t. Buy the second one,” he says. “The clients who get real value from AI didn’t set out to buy AI – they set out to fix an outcome, and using AI turned out to be the best way to get the job done”:  

According to Mackean, starting with the outcome, rather than with the technology itself, forces the right questions. What does success mean? Who will be using it? How can the process be simplified? This outcome-first approach shapes the rest of the process.

Once you know the outcome, you can audit the data that will feed the decision and apply the proper governance. In large real estate organizations, different teams work with different clients, assets, and tenants, all of whom have legitimate expectations about how their information is handled. “Sometimes you have specific functions that need certain data, but you don’t want them to have access to everything,” Mackean says. “By segregating agents, you can make sure they only have access to certain information. You want AI to be trained, but you don’t want it to learn across different clients.” In other words, implementing the right guardrails early on can prevent access to sensitive data while recording a living history of who has accessed what.

Integration is the next consideration. Most real estate tech stacks are a collection of platforms built independently and connected imperfectly. Property management systems, lease management platforms, CRMs, and market data providers all hold information that AI needs, and getting them to share it reliably takes ongoing work rather than a one-time setup. “‘It worked in the pilot’ isn’t a readiness claim. What’s your plan for month six, after the model, the data and the portfolio have all moved?”  Mackean noted. 

An integration or model that works today can break quietly following an update that changes how data is structured or interpreted. “Sometimes nobody notices until an AI output starts degrading in ways that are hard to trace. One example was a client with an office utilisation agent that over a few weeks showed an unexplained increase in utilisation across all buildings by 15% after a change in hybrid-work policy that should have reduced utilisation. After we investigated we saw it was due to an upgraded badging system for security reasons that had more advanced tailgating logic, rollout of additional wifi access points from totally independent workstreams but had significant impacts for the company’s portfolio strategy. “

Because the models themselves are changing so quickly, it also makes sense to build integrations with more than one of the major AI providers. The models available today are considerably more capable than those from two years ago, and that trajectory is not slowing. “Avoid lock-in with a specific vendor or partner because it helps you stay flexible as the models change,” Mackean said. An architecture that lets you swap the underlying model without rebuilding everything around it costs little to maintain and is worth a lot when the next generation arrives.

Some thought also needs to be put into which models are best for which tasks and how advanced a model the task really needs. Models range from lightweight and cheap to sophisticated and expensive, and using the top tier for everything gets costly fast. “If you are not careful you can find that 90 percent of staff are using the highest level of AI, costing significant sums of money, on tasks where it might not be needed,” Mackean said. Deciding in advance which workflows warrant which model tier, and giving staff clear guidance on that, keeps costs manageable as usage scales.

All of this work happens before a single AI tool goes live, which is exactly the point. The readiness gap is not a technology problem, and it does not close on its own. It closes when an organization has done the work of defining outcomes clearly, auditing its data honestly, mapping the integrations that will need ongoing attention, and building in the flexibility to adapt as the models keep improving. That work is difficult and time consuming, particularly for organizations that have never had to think about their operations at this level of granularity. This is why a growing number of real estate companies are bringing in outside advisors to help them assess where they actually stand before committing to a deployment path. 

So as tempting as it is to be one of those companies “actually integrating AI,” there is real value in taking the “thinking about it” stage seriously. The organizations that spend that time productively, applying the right framework and asking the uncomfortable questions early, are the ones whose pilots produce results worth scaling. The ones that skip it tend to discover the gap later, after the budget has been spent.

The post The Questions Real Estate Organizations Should Answer Before Deploying AI appeared first on Propmodo.

​