Getting your data house in order used to be something you'd tackle eventually, you know, right after fixing all those other pressing issues on your list. But here's the thing - if your data isn't ready, those shiny new AI tools you're hearing about basically become expensive paperweights. L.D. Salmanson, co-founder of Cherre, spent years building the infrastructure that connects fragmented real estate data, and now he's watching something pretty remarkable happen as AI agents join the party.
"We launched in 2018 with the explicit vision of connecting real estate data to enable our clients to deliver advanced AI solutions," Salmanson explains. "We wrote this memo which said whoever wins data, wins AI."
That memo turned out to be rather prophetic.
The Three-Layer Problem Nobody Wants to Talk About
Real estate companies all have basically the same three data headaches, according to Salmanson. First, they need access to information that's scattered everywhere - in their own systems, in partner systems they don't control directly, or bought from various vendors. Second, they need all that data to actually speak to each other. One system calls something an address, another calls it a parcel ID, and yet another has the same address spelled differently. Third, they need that connected data to turn into something business users can actually do something with.
"At the core of that problem was first connecting data," Salmanson notes. For Cherre's clients - major asset managers, banks, and insurance companies - that means connecting internal operational and financial systems with third-party vendor data through over 50 partner networks. This is the role of Cherre CONNECT, the integration layer that securely ingests and standardizes fragmented information into a unified pipeline.
From there, data flows through Cherre QUALITY, the control layer that allows clients to see every single thing that runs through the data pipeline. We're talking observability, business logic rules, SOC 1 and SOC 2 compliance audibility - all the things that really matter when you're managing billions in assets.
Once data connects to Cherre's core module - aptly named, Cherre CORE it is mapped into a universal and semantic data model, anchored by a knowledge graph. That knowledge graph? It's not small - literally billions of nodes representing over four billion legal entities.
Once data is connected, governed, and modeled inside CORE, intelligence becomes operational. That intelligence layer is what Cherre calls ALPHA - where unified portfolio data translates into acquisition analysis, underwriting insight, and strategic decision support across asset classes.
With Cherre ALPHA in place, AI agents are no longer querying disconnected systems. They’re reasoning over structured, contextualized intelligence - which is where scale becomes possible.
Why Data Readiness Isn't Optional Anymore
Having spent decades in content strategy and tech before founding Business Intelligence Group, I've watched plenty of technology waves come and go. The companies that surfed those waves successfully were usually the ones who did the boring infrastructure work first. Real estate data management is basically that boring work that makes the exciting stuff possible.
"Our clients are very creative," Salmanson observes. Historically, Cherre helped them get data out of various sources and put it reliably in places where clients could access it. That mostly looked like reporting - dashboards showing what happened in the past or what's happening right now. Data scientists would also build products on top of that connected data in their own environments.
But something changed when AI agents entered the picture. According to Salmanson, "What we've enabled with Cherre’s Agent STUDIO is allowing business users to be able to interact with that data in a way that goes well beyond what we could have done before."
Business users - not just data scientists - can now interact with complex data in natural language. They can ask questions, get answers, and even have the system reason about things that weren't pre-programmed. The difference between this and older BI tools is actually pretty stark.
From College Intern to Diagnostic Doctor in One Year
The evolution of AI capability is accelerating in ways that honestly feel surreal. A year ago, you might explain AI as similar to a fourth grader - it could do tasks but would come back with errors you'd need to fix. Six months ago, it jumped to post-college graduate level, basically capable right out of the box.
"It's still an intern," Salmanson quips when I suggest it's at college graduate level. Fair point - it still needs supervision. But the trajectory is clear.
"I you think about the industrial revolution, it decoupled a unit of physical work with a unit of physical human," Salmanson reflects. “That revolutionized everything. The internet removed the need for humans to remember or manually look up information. What we're seeing today is a very fast decoupling of that last mile of reasoning."
All the information is at your disposal, so now the question becomes: what do you want to reason about and why?
The Agent Marketplace Approach Nobody Expected
Cherre's taking an interesting approach with their Agent Marketplace. Rather than building everything in-house, they're creating what Salmanson describes as something like the GitHub of data and AI agents. Companies can share knowledge and pre-built solutions rather than everybody rebuilding the same wheel.
"We want to allow our clients to actually be able to share their knowledge in the same way they're able to share and consume third-party data today," Salmanson explains. If someone at Blackstone builds an agent that evaluates ESG compliance for a property, and JP Morgan needs something similar, why should JP Morgan start from scratch?
Agent Marketplace lets companies package their agents, share them (either openly or for compensation), and build on each other's work. Some agents will be private, some public, some monetized. The model makes sense because it reduces redundant compute power and accelerates innovation.
"The models don't want to repeat the wheel either," as Salmanson puts it. More compute power gets expensive rather quickly.
What Happens When Humans Step Back
The role of humans in this new landscape is shifting into something more abstract. You're no longer doing the reasoning - you're deciding what's worth reasoning about. You're arbitrating between different AI outputs that might both be "true" in different contexts.
Cherre uses the term "epistemic arbitration" to describe this. The notion of truth becomes fluid in these systems because truth gets defined differently depending on your semantic model or context. Both might be true for their specific purposes, but you need humans to decide which truth applies to the decision at hand.
"What's interesting to explore and then letting the models go explore that. So I think we're well on that journey."
The models are getting scary good at diagnostics and analysis. What they're weak at - where humans still fill critical gaps - are those situations where outcomes become unpredictable or where judgment requires weighing factors the AI can't fully grasp.
The Practical Reality Right Now
Let's be honest about where we actually are today. These AI agents aren't magic. They cost money to run at scale. They make mistakes. They need guardrails. They require data that's been properly connected and structured.
That last point is probably the most important one. Companies jumping straight to AI implementation without fixing their data infrastructure are basically building on quicksand. Cherre's entire business model validates this - you need the boring plumbing work done first.
For real estate firms specifically, this means connecting internal, financial, and operational data, with third-party market data, and vendor information into a single source of truth. It means establishing reliable, repeatable processes with proper observability. It means building knowledge graphs that understand relationships between entities.
Only then do the AI agents become genuinely useful. Without that foundation, you're just getting confidently wrong answers delivered very quickly.
The good news? Once you have that foundation, the capabilities expand remarkably fast. Agents can analyze properties, evaluate market conditions, assess risks, and support decisions across the portfolio. They can work 24/7, scale instantly, and get smarter with every interaction.
Business users get answers to complex questions without waiting for data scientists to build custom reports. Decisions that took weeks now happen in minutes. The time and cost reduction is significant, which is why Cherre's clients keep finding creative new ways to deploy agents across their operations.
This isn't really about replacing humans. It's about letting humans focus on the questions that matter rather than the mechanical work of finding and analyzing data. It's about making expertise scalable. It's about turning data from a compliance burden into a competitive advantage.
As Salmanson notes, we're well into this journey already. The models keep getting smarter. The infrastructure keeps getting better. The applications keep getting more sophisticated. This is as dumb as AI is going to get, and it only accelerates from here.
For real estate companies still treating data management as something to tackle eventually? Eventually just arrived, and it's bringing some powerful new friends along for the ride.