The AI Advantage in CRE Is Not the Tool. It Is the Workflow You Own.

A junior analyst spends most of their first two years doing one thing: taking a seller's offering memo and re-running every number against the firm's own assumptions. The vacancy that reads as "stabilized," the generous exit cap, the capex that quietly disappears from the pro forma. As Thesis Driven put it in a recent workshop, converting a sales document into an honest internal memo is exactly the kind of work an AI agent now does in minutes.

That sentence should stop every asset manager cold, but not for the reason most people assume. The headline is not that AI can screen a deal faster. The headline is a question hiding underneath it: when that agent runs your acquisitions logic, whose logic is it, and where does it live?

This is the real inflection point in commercial real estate right now. The market has moved past "should we use AI." The serious owners are wiring AI directly into the workflows that define their edge. And the firms that control the data, the workflow logic, the permissions, and the operating context behind that AI are pulling away from the firms that rent all of it from a vendor.

The tool was never the hard part

Walk any CRE conference floor in 2026 and you will find hundreds of AI tools. ChatGPT, Claude, and a growing shelf of point solutions that promise to summarize leases, draft memos, and flag risk. The tools are not the constraint. As Best Ever CRE argued recently, most real estate professionals do not have an AI-tool problem, they have an implementation problem.

I would take that one step further. Implementation is not really about which model you pick. When AI capability converges, and it is converging fast, model selection becomes procurement, not strategy. What actually differentiates one portfolio from another is four things only the owner can build: proprietary data, the operating workflows that turn that data into decisions, the orchestration layer that connects them, and the institutional knowledge encoded into those systems.

An AI agent is only as good as the context you feed it and the logic you wrap around it. Give a generic model your acquisitions criteria, your underwriting standards, your operating history, and your renewal patterns, and it becomes genuinely useful. But that usefulness sits on top of your data and your logic. If those assets live inside someone else's platform, you have not built an advantage. You have rented one, and the meter is running.

There is a subtler cost too. When your workflow logic lives inside a vendor's tool, you lose the ability to see how a decision was reached. The agent flags a deal, or passes on one, and the reasoning is a black box you did not build and cannot inspect. That is fine until an investment committee asks you to defend the pass, or a lender wants to understand how you underwrote the renewal assumptions in a refinancing package. Ownership of the logic is also ownership of the explanation.

Watch what the serious owners are actually building

The pattern is easiest to read at the top of the market. The largest owners are not waiting for a vendor to package intelligence for them. As Commercial Observer reported, Blackstone and Brookfield are using AI to build real estate technology in-house rather than buying it off the shelf. Fisher Brothers, an owner rather than a technology firm, launched its own AI innovation lab to develop capabilities against its own portfolio.

Read those moves for what they signal. These are sophisticated owners deciding that the intelligence layer is too strategic to outsource. They are not doing this because building in-house is cheap or easy. They are doing it because they understand that the workflow logic, once encoded, becomes a durable asset. Every deal screened, every memo drafted, every operating pattern captured makes the next decision sharper. That compounding only happens if you control the system doing the learning.

Meanwhile, the data providers are moving to meet AI where it lives. Trepp is demonstrating how to connect trusted CRE data to AI, and Green Street has introduced institutional-grade intelligence you can query directly. This is a healthy development. But notice the structure it creates. If your proprietary operating data never joins that conversation, your AI reasons entirely on market averages and vendor-supplied context. It knows the market. It does not know your building. That gap is exactly where owner advantage is won or lost.

The economics hide until diligence finds them

Here is the part that belongs on an asset manager's desk, not an IT roadmap. The cost of renting your intelligence does not show up as a line item. It shows up as a decision you could not make because the data was not yours to reach.

Consider the acquisitions workflow again. If your screening agent runs inside a vendor platform, three things are true whether you like them or not. First, the logic you refine improves the vendor's product as much as yours. Second, you cannot easily audit why the agent flagged or passed on a deal, which is a problem when the investment committee asks. Third, if you leave that vendor, the accumulated tuning leaves with them. You start over at a new address, on a new platform, with none of the learning you paid to build.

Now extend that across the operating side. Actual tenant consumption data, lease-level performance, equipment run histories, the causal drivers behind your OpEx variance. When that data is trapped in disconnected vendor systems, your AI cannot see the whole picture, and neither can your diligence team at refinancing or exit. A buyer's advisors will find the gaps. When operating history is thin or unverifiable, the price moves against you before you sit down at the table.

The math here is not abstract. On a multi-tenant office asset, the operating improvements that owner-controlled data makes visible run in the range of $0.60 to $0.90 per rentable square foot per year. Capitalize that recovered NOI at prevailing cap rates and every dollar becomes roughly fifteen to twenty-five dollars of asset value at refinancing or exit. That is the swing sitting inside a data question most owners have never framed as a data question.

This is the quiet math behind a simple truth. If you don't own your data & digital infrastructure, your vendors do. And when the intelligence that runs your portfolio is built on rented ground, every improvement you make is partly an improvement you are handing to someone else.

Owner-controlled intelligence is a structural choice

The way out is not another tool. It is a foundation you own and a governed layer that lets any AI operate under your permissions rather than a vendor's.

This is the OpticWise thesis, and it is built as two layers. The first is managed data & digital infrastructure: the owner-owned foundation that consolidates building connectivity onto a single, secure, segmented base through BoT® (Building of Things®). This is where your data gets captured, normalized, and made portable instead of scattered across systems you cannot reach. It delivers value on its own, and it does not tax your on-site engineers or property teams, because governing data & digital infrastructure is a different discipline than running a building.

The second layer is the owner-controlled intelligence layer, Property Brain™, which scales into Portfolio Brain™. This is a vendor- and LLM-agnostic layer: a governed data plane plus a trust plane. It means you can plug in any decision engine, any AI model, any vendor analytics tool, and each one acts under your governance, on your data, with full lineage and permissions intact. Swap models when a better one arrives. Swap vendors without losing your history. The intelligence stays yours.

The path to get there is the PPP 5C™ plan, drawn from Peak Property Performance®. Clarify: run a review to define success metrics, map who owns what data today, and identify where it leaks. Connect: establish secure, owner-controlled connectivity that repeats property to property. Collect: capture and normalize high-fidelity data into a consistent model you can reuse. Coordinate: govern identity, access, privacy, lineage, and rules of use. Control: let any AI or workflow act on that data under your permissions. Steps one through three build the foundation. Steps four and five turn it into intelligence that compounds.

Where this leaves the asset manager

The owners building AI into their workflows right now are making a bet that intelligence is a compounding asset, and that whoever controls the data and logic underneath it controls the return. I think that bet is correct, and I think the window to make it deliberately, rather than reactively, is open now.

Here is the practical starting move. Pick one workflow that matters, acquisitions screening or renewal analysis are good candidates, and ask three questions. Where does the data that feeds it live? Who can reach it, export it, and reuse it if you change vendors? And is the logic you are refining accumulating as your asset or someone else's? If you cannot answer cleanly, you are renting intelligence, and the compounding is happening on the wrong balance sheet.

That is the conversation worth having before the next model release makes the tool question feel urgent and the ownership question get skipped again. The tool will keep getting better and cheaper for everyone. What you own underneath it is the only part that stays yours.

Own your data & digital infrastructure. Operate with strategic foresight. Build for the long game.