Two announcements this month show two ways owners could end up using AI, and the next AI contract you sign will move you toward one of them. Thesis Driven reported that EliseAI, a company known for AI leasing and resident communication, with roughly 6.5 million units on its platform, is building its own AI-first property management system. Around the same time, eXp Realty told its agents they can connect ChatGPT, Claude or Gemini directly to the brokerage's data.
The question I'd put in front of every owner right now is where your AI tool will read its data from, and who decides what happens to that data when the contract ends.
Two ways AI is arriving in real estate
If EliseAI's system ships as described, one company would run leasing conversations and renewals and would also hold the system of record: rent rolls, lease terms and resident history.
eXp took a different route. According to a Real Estate Solutionist roundup of eXp's "AI of Choice" strategy, agents can connect their preferred AI assistant to authorized brokerage data through the Model Context Protocol, an open standard for linking AI tools to data sources. An agent can pull production figures, leads, transactions and CRM records into the assistant they already use. The brokerage decides what any assistant is allowed to see, and each agent chooses the assistant.
Both setups can produce useful AI. What matters to you as an owner is where the records sit and who writes the rules for them. When the records sit with the business and AI tools connect under the business's permissions, you can change tools and keep everything underneath. You feel the effect of that choice years later, when you renew, refinance or sell.
Owning your data and controlling it are different things
Owners tell me all the time that they own their data, and the contract usually says so. The trouble starts the day you try to use it.
Mr Family Office draws this same line between owning data and controlling it, and in a building, control comes down to five practical tests. Can you export the full history in a usable format whenever you want it? Do you hold the admin credentials yourself? Do you set who and what can read the data? Do you keep the history when a vendor leaves? Can you connect a different tool to the same data next year without rebuilding?
Every test you fail makes the contract language worth less, because the data still has to get from the vendor to whoever needs it: a lender, a buyer, or the next AI tool you want to try.
Third-party management makes this harder, and it's worth checking. In many arrangements, the management firm holds the contract with the software vendor. You answer for the investment, yet you may have no direct relationship with the company storing your operating data, and your access depends on what the manager's contract allows. If you change managers, that data relationship may stay with the old firm. Put an AI tool on top of that system and you've added one more dependency you don't control.
If you don't own your data & digital infrastructure, your vendors do.
What a bundled AI platform could cost you over a hold
Think about it the way you'd think about building out a suite in someone else's building. The improvements are real and they help you every day you're there. When the lease ends, they stay with the landlord.
An AI that comes bundled inside a vendor's system of record can work the same way. Your team builds prompts, automated workflows, saved reports and renewal playbooks inside it. If the platform connects only to the systems the vendor chooses to support, your meter readings, access control logs and building management data may never reach the AI at all. Over a few years, a growing share of how your team works would live inside one company's product. Leaving would mean rebuilding those workflows, retraining staff, and accepting whatever export the vendor provides. Both sides know that at renewal, and it tends to show up in the price.
You pay that price out of operating expenses, and operating expenses come straight out of NOI. The arithmetic is simple: divide any change in annual NOI by your cap rate and you get its effect on asset value, so at a 6 cap every dollar of NOI moves value by more than sixteen times its size. A cost that ratchets up at each renewal is a cost you carry into your valuation.
The second cost comes when you need to explain your operating results to someone outside the company. A lender reviewing a refinance, or a buyer's diligence team, wants the numbers behind your NOI and evidence they hold up. If the AI's outputs and the data behind them sit somewhere you can't export from cleanly, your team has to piece the package together by hand, and you'll show less than you actually know about the building.
The third cost is buying it twice. An AI setup tied to one property's vendor stack would have to be bought and configured again at a property that runs different systems, and most portfolios mix systems from building to building.
The same thing happens with your firm's judgment. Thesis Driven describes firms writing their acquisition and investor-reporting standards into Claude skills, so the standards are written down once and applied the same way no matter who runs the analysis. When those standards live in a format your firm controls, they stay with you through a manager change or a vendor change. If they're built into one vendor's proprietary tool, they leave when the vendor does.
Build the layer, then choose the AI
At OpticWise we build an owner-controlled layer in two parts.
The first is managed data & digital infrastructure: the connectivity and systems in your buildings, designed to a repeatable standard, documented, and operated for you, so your property managers and engineers can stay focused on running the building.
The second is an intelligence layer that collects data from systems such as your property management system, meters, access control and building management, under your credentials and your rules. We call it Property Brain™. It's vendor- and LLM-agnostic, meaning an AI from any company can work on that data within the permissions you set. Once you've standardized it at one property, the same model, rules and connections can be reused at the next, which is how Property Brain™ becomes Portfolio Brain™.
We get there in five steps we call the PPP 5C™ plan.
Clarify comes first. We map which data exists, where it lives, whose contract it sits under, and what you can export today. We call this step a PPP Review.
Connect sets up secure, owner-controlled connectivity that works the same way from property to property.
Collect captures and normalizes data from the systems you choose into one consistent model you can reuse.
Coordinate governs identity, access, privacy, lineage, retention, and rules of use.
Control is where AI comes in. A vendor platform, your own analytics, or any LLM your team prefers can act on that data under your permissions. If a better AI comes along in two years, you could connect it to the same layer and keep the history, rules and workflows you've already built.
What to do before your next AI contract
"You're writing the check. You own the roadmap," as I said on the Peak Property Performance® podcast (Ep. 32). With AI, the roadmap you accept this year gets harder to change every month your team builds on it.
Before you sign or renew anything with AI in it, get written answers to four questions.
First, can we export the full history, including what the AI generated and the data it used, in a standard format, on demand, at no extra charge?
Second, who holds the admin credentials, and will we hold them ourselves?
Third, if this relationship ends, do we keep the data, the workflows, the configurations and the history?
Fourth, can this AI read data from systems outside the vendor's platform, and can a different AI read the data stored inside it?
If the vendor can't answer those clearly, hold off on signing until they can. If your management firm holds the software contract, read what your management agreement says about data access well ahead of the renewal date.
Then pick one asset and start with a PPP Review. We'll map who holds which data, where you lack control, and a practical path to an owner-controlled layer any AI can connect to. If the case doesn't hold up for that building, you keep the map and decide from there.
Own your data & digital infrastructure. Operate with strategic foresight. Build for the long game.
