CRE’s AI Problem Isn’t the Application Layer
I've spent 20 years as a product leader in the tech space, with deep expertise in building and scaling AI products at Google, Two Sigma, and Point72, where I’ve refined skills that are the intersection of product and data excellence. Stepping into the role of VTS' new Chief Product Officer, I'm energized by the opportunity in the commercial real estate industry to help professionals not just use AI, but harness its power at the highest potential, and truly trust it.
Trust starts with education. “Big data” can sound abstract, even intimidating, but at its core, it’s about transforming millions of fragmented data points: market comps, lease terms, capital flows and more into clear, actionable insights. Today’s data engines don’t just aggregate information; they detect patterns and trends, surface risks and opportunities, and make sense of information that’s disparate and would otherwise remain unutilized. When professionals understand how these systems work, how models are trained, how outputs are validated, and where human oversight fits in, confidence grows.
For years, CRE has evaluated AI as though better models would unlock better operations. In reality, the constraint was never cognition, it is coordination. Every asset is governed by a web of inputs that exist in fragmented systems with no centralization, and while AI operating in that environment may produce answers, it cannot deliver meaningful outcomes. The truth is, the industry hasn’t lacked intelligence; it has lacked a unified memory, and the moment those sources align into a continuous operational picture is the moment AI stops assisting work and starts shaping it.
Every building produces information continuously: leasing activity, legal agreements, operating obligations, tenant relationships, financial performance. But those inputs live in different systems, are managed across different teams, move at different speeds, and rarely reconcile into a single operational reality. AI cannot be asked to reason on top of that fragmentation.
The industry’s starting point was to improve interpretation. Better extraction. Better summaries. Better conversational answers. But the key to operating buildings most-effectively is making sense of all the data available and understanding what is fundamentally true and what needs to happen next.
Leases Are Not Documents, They Are Behavioral Rules
A lease is often treated as a source of data. In practice, it is the DNA governing how an asset evolves over time.
Rent escalations change revenue.
Notice periods trigger workflows.
Exclusivity clauses affect leasing decisions.
Operating obligations affect compliance.
These are not static facts to retrieve. They are the inputs of an ever-expanding context window influencing operations.
Most systems, however, treat the lease as an artifact that periodically produces structured outputs, an abstract, a report, a set of fields. The moment activity occurs, the system drifts away from the contract it is supposed to represent. Teams then reconcile differences manually: document vs spreadsheet vs system vs reality.
AI layered on top of this environment becomes advisory by definition. It can describe the lease, but it cannot reliably determine what applies right now across the portfolio.
The breakthrough capability is not understanding, it is maintaining clarity over continually-evolving context.
Intelligence Requires a Unified Understanding
In operational industries, software eventually converges on the same requirement: it must maintain a living representation of the world, not periodically describe it.
For commercial real estate, that means connecting the full lifecycle: proposals, executed leases, amendments, renewals, obligations, performance, into a single operational state that persists over time.
When systems remain disconnected, every team reconstructs reality independently. Leasing interprets the agreement one way, asset management another, accounting a third. AI then reasons over whichever version it encounters, producing answers that appear correct but are operationally unsafe.
When the state is shared, the role of AI changes completely. The question stops being “what does the lease say?” and becomes “what should occur now?” Not because the model improved, but because the environment contained connected data that delivered actionable insights.
The Shift From Information to Infrastructure
The industry often frames AI adoption as a model choice. In practice, it is an infrastructural choice.
If lease data functions like documentation, AI produces explanations.
If lease data functions like infrastructure, AI produces dependable guidance.
The difference is data competency, the ability of the platform to keep multiple sources synchronized into a usable operational context and deliver that context consistently to both people and software.
This is why the next generation of CRE platforms will be differentiated less by features and more by whether their data can act as a trusted operational reference point.
The goal is not software that occasionally interprets buildings, it is software that continuously understands them. Once that foundation exists, intelligence is measured less by how well the system answers questions, and more by how rarely teams need to ask them.
