The Importance of a Unified Language for AI Development in CRE

As the technology driving artificial intelligence evolves at light speed, so too does the need for commercial real estate companies to learn, adapt, and integrate that technology to ensure they’re optimizing their workflows and not falling behind their competition.

Raj Singh, CEO of tech provider Altrio, emphasized on a recent Commercial Observer webinar just how crucial this was for all within CRE.

“Real estate is a services industry, and most of the value firms add comes from people sitting at computers, processing information, and making decisions. AI goes right to the center of how these firms create value,” said Singh. “Intelligence and mental bandwidth have always been a major constraint. Now, they can be added quickly and relatively cheaply, practically without limit. So as that constraint loosens, the CRE professional’s job shifts from doing the work to directing it — from playing an instrument to conducting an orchestra.”

The November 10 webinar was presented by Altrio and hosted by Edward Cohen, director of technology and AI partnerships and host of the Tangent podcast at Commercial Observer, on behalf of Commercial Observer Partner Insights. It featured an in-depth conversation about how the real estate industry is optimizing AI agents for scalable and repeatable results between Singh and L.D. Salmanson, senior vice president of real estate data intelligence platform Cherre at RealPage. 

(Cherre, for which Salmanson is co-founder, was acquired by RealPage in July 2026.)

One essential consideration for CRE firms in integrating AI processes into their workflows is the semantic layer, which Singh defined simply as the need for firms to write down definitions to ensure that teams have identical references as they scale.    

“People can be befuddled by terminology,” said Singh. “If you’re leading a team of two or three people and you have plans to scale that team to 15 or 20 people, you need to start writing things down. The shared understanding that might be easy to maintain when you’ve got three people working on something becomes much harder when you’ve got 20 or 30. So you need to start agreeing on what terms mean and how your processes are meant to run.”

Singh also advised companies to realize that the introduction of AI into everyday workflows should itself been seen as a form of scaling.   

“In going from having a few people to having a few people and all their [AI] agents, even small groups within a large company need to start thinking about themselves as bigger groups within that company. That drives the complexity challenges.”

For those less familiar with semantic layers, also known as ontologies, Salmanson noted that we can often have different definitions of common terms.

“It’s often the case that we have different meanings for things within an organization for good reasons,” said Salmanson. “For example, we calculate net effective square footage differently in Chicago than in New York. I don’t know why. But if we use the same definition of net effective square footage but mean something else, our calculations will be skewed even thought the underlying data hasn’t changed. So we need to know that.”

And by “we,” in this case, he means not only the people involved, but any large language models (LLM) that companies might be using to process and rely on this information. 

Singh then turned to the issue of prioritization within reconsiderations of the tech stack, and the importance of staying focused on company goals. 

“Everyone I talk to in the industry is rethinking their tech stack, and how they should leverage technology to maximum effect,” said Singh. “Because it is so much easier to build things, it forces you to work harder to stay focused.”

He cautioned companies against succumbing to the “dopamine effect” AI creation can bring, recalling a LinkedIn post by a CEO who said he had spent a few weekends vibe coding, and was so enthused about the results that he soon realized he’d been neglecting his day job. 

“This stuff is so rewarding when you’re able to solve problems and build things,” said Singh. “But it forces you to focus on what you should build. Just because it’s easier and cheaper to build something doesn’t mean the cost is zero. Prioritization still matters.”

Part of this is determining what you should build verses what you should buy.

“The bar to build software has never been lower,” said Salmanson. “The bar to build good software that’s differentiated and worthwhile has never been higher.”

To this end, Salmanson said he encourages companies to “map out all the tasks and information flows within an organization,” and then optimally identify software from an expert vendor that can address the scenario from all angles. 

“If you think it’s a problem that nobody’s solving right now and you need it, then fine, build it,” said Salmanson. “But the second somebody offers it, you should kill that solution — meaning, know that it’s a temporary solution until something better comes along, because you’re not a software company.”

The only item Salmanson encourages clients to build internally is data infrastructure, but even then, he recommends doing so with experienced partners.  

“Firms who think they’re saving money by building on their own are losing by not getting the infrastructure they need to win in the future,” said Salmanson.

Singh added that when it comes to technological learning curves, the curve for AI might be the steepest he’s ever seen, leaving yet one more reason for companies to leave AI build to those with the luxury of working on it 24/7. 

“The speed of [AI progress] does not leave any time for any real learning,” said Singh. “This is almost moving too quickly for folks to make mistakes on that they can learn from.”