The commercial real estate industry is deep into its first real AI cycle. Vendors are multiplying, demos are everywhere, and board-level pressure to have an AI strategy is at an all-time high. Most operators are asking the same question: what does the right CRE technology stack actually look like? The short answer: financial infrastructure, an operating layer that spans leasing and asset management, and AI built on top of a structured data foundation — not instead of the systems already in place.
Here’s how to think through each layer.
The financial backbone: why your ERP stays
If you run a portfolio, you likely have an ERP. MRI, Yardi, JD Edwards, etc. It closes the books, generates the rent invoices, tracks the ledger, produces the reports your accounting team lives in. It is, by any measure, mission-critical infrastructure — and the idea of replacing it wholesale is a fantasy no serious operator entertains.
This system is not going anywhere, nor should it. The system of record for billing should be conservative, durable, and hard to change.
The operating layer: where the business runs
An ERP records what happened to the money. But an enormous amount of the value in commercial real estate is created before a dollar ever hits the ledger — in the deals being pursued, the relationships being managed, the decisions being made about what to lease, to whom, on what terms, and how to maximize returns over time.
VTS started here, in leasing pipeline management. Over the last decade, the VTS platform became the system where the industry’s leasing activity is executed and recorded: the pipeline, the proposals, the deal economics, the tenant relationships.
But leasing is only part of the picture. It’s the part we’re best known for and where we’ve grown from in recent years. The harder, messier work is everything that comes after the deal closes: managing the asset itself.
- Which leases are rolling in the next 24 months?
- Where are renewal risks concentrated?
- Which co-tenancy clauses are live right now?
- Which landlord obligations are coming due?
None of these are leasing questions, exactly — they span legal, finance, operations, and leasing all at once, which is precisely why they’re so hard to answer. For most of the industry’s history nobody could answer them completely, because the data lived in documents scattered across different systems and different teams.
That’s the problem VTS moved into: not just running the deal, but managing the asset through every amendment, renewal, and obligation that follows. We do that by ingesting, abstracting and structuring the source documents. And because those source documents now live inside the operating layer, VTS becomes a second set of eyes on what’s in the ERP — helping reconcile what the lease actually says against what’s being billed, so discrepancies surface before they become write-offs or tenant disputes.
Do all of that work across a portfolio, year after year, and it produces something with its own compounding value: data. And running a modern portfolio well takes two kinds of it, despite most operators only thinking about it as one.
The data advantages: portfolio intelligence and market intelligence
The first is the data you generate inside your own portfolio: your leases, your amendments, your deals, your obligations, your entire document history. Getting it into a form you can actually use is the most important part of the stack, and the most consequential to get right. It isn’t a folder of files to search. It’s a web of interconnected, relational data, where an amendment signed three years ago quietly modifies an option that changes an obligation that shifts your exposure today. Building the intelligence layer means making sense of that tangle: structuring every document, connecting it to every other, so each clause understands what it affects and what affects it. That structured foundation is what unlocks meaningful intelligence, everything useful downstream depends on it.
Because VTS sits at the center of so much leasing activity, we see demand signals that aren’t available anywhere else. Which tenants are actively touring. Which buildings are seeing requirement velocity pick up. Which markets are 6-9 months away from a move that won’t show up in traditional metrics until it’s already happened. That market view lives in VTS Data — and it’s the difference between making decisions with lagging indicators and making them with a real forward view of where demand is going.
Put the two together and you get something no single system has offered before: a genuinely holistic view. Look backwards with confidence — understanding exactly what your contracts commit you to and precisely where your exposure sits — and look forward with that same confidence, knowing where demand is actually headed.
This is why the data foundation has to live inside the operating layer, not alongside it. The intelligence is only as good as the system underneath it — the one where deals are created, documents are generated, and leasing activity is recorded in real time. Separate the data from the operating layer and you’re back to reconciling two systems manually. Keep them together and the intelligence compounds automatically.
What to look for when evaluating AI for commercial real estate
Once that stack is in place — financial infrastructure, an operating layer that spans leasing and asset management, and a data foundation that connects your portfolio to the broader market — the question becomes: what should you actually look for when AI is being added on top of it? Because not all of it is the same, and the differences matter more than most evaluations surface.
A few questions worth asking of any AI in this category:
Questions:
- Was it trained on real leasing transactions, or just documents?
A model that has never seen a deal negotiated cannot help you negotiate one. Generic document AI and purpose-built leasing AI are different products. Ask the vendor directly. - Does it understand a lease as a living document?
The LOI, the executed lease, the three amendments, the estoppel that quietly modified an option — these are connected. A system that treats each as a standalone artifact is one amendment away from giving your team wrong information. - Can you query your entire portfolio at once?
Understanding renewal exposure across fifty leases should take one question, not fifty. If the answer is one lease at a time, that’s a search tool. - Is every answer traceable to source?
“The system says so” stops being an answer the moment a tenant challenges a charge or an auditor asks where a number came from. Every data point should be traceable to the exact clause in the exact document it came from. - Does it create, or only describe?
AI that summarizes what already happened is useful. AI that drafts a proposal from an LOI in seconds — with your portfolio standards and deal precedent pulled in automatically — is a different category of value. Ask to see both. - Does it get better as your portfolio grows?
General-purpose AI pointed at a folder of PDFs works reasonably well on one lease. Across hundreds, it degrades — re-reading and re-deriving the same facts from unstructured text on every query is slow, expensive, and increasingly unreliable at scale. Purpose-built systems structure each document once and answer against that structured layer. The advantage grows with the portfolio.
The teams pulling ahead right now are not the ones who bought the most AI tools
They’re the ones who got serious about what their stack needed to look like first:
- Financial infrastructure that owns the ledger
- An operating layer that connects leasing, asset management, and the full document history of every asset — with the data foundation living inside it, not alongside it, so the intelligence compounds instead of fragmenting
- AI built on top of that foundation, not pointed at a folder and called a strategy
Get those right, and AI starts to compound in your favor.
Want to see what this looks like in practice? Talk to our team.
