Your building's repair history, vendor records, and access logs are about to do more work than they ever have. Software is starting to use them to act: triaging maintenance requests, dispatching vendors, and routing invoices for approval. That raises the stakes on a question most owners haven't looked at closely. Who holds that history, and do you keep it when the platform changes hands?
At its Ascend 2026 conference, MRI Software showed how its Agora platform is being developed to coordinate AI agents and automate real estate workflows across the software it already sells. As Realcomm put it, the next phase of AI "may be less about answering questions and more about getting work done." Some operators are already seeing results. California property manager Dream Big reports that repairs fell from 14 days to four after it automated more than 90% of maintenance coordination with an AI maintenance tool.
Cutting a repair from fourteen days to four means fewer frustrated residents and a better shot at the renewal. Getting there depends on records you may not control.
What an agent needs before it can do anything
An AI agent needs three things before it can do useful work: it has to reach the data, the data has to be in a shape it can use, and it needs permission to act. A Thesis Driven deep dive on a company building a data layer for property software makes the first two points plainly. Agents work only when the data sits where they can reach it and in a usable form.
A maintenance agent that sends a plumber at night needs several records to get it right. It needs the unit's repair history, so it can tell a recurring leak from a new one. It needs the approved vendor list and how each vendor has performed. It needs access records, so the plumber can get into the unit. It needs the approval limits that decide when a person has to sign off. Each of those records lives in some system, and the agent can use only what that system lets it see.
When those records are spread across platforms, every property becomes its own project. One building keeps work orders in one system, access control in another, and vendor invoices in a third. An agent built into one platform can act on that platform's data and can't reliably reach the rest. It also can't tell you whether a roof repair went better at your Denver property than at your Phoenix property if the two run on different software.
Your operating history is now worth something to other people
Once software can act on operating data, other people want that data. I think platforms will build their agent features around the history they already hold, because the more of your records they have, the more of your building's work they can automate. That makes your history part of their pitch, to you and to anyone who might buy them.
A more direct market is also forming. Peter Lohmann has written about AI data labs approaching property managers to buy their data. Work orders, resident messages, and vendor records make useful training material for someone else's product. Whether a manager can sell that data, and whether you get a say, comes down to contract language few owners have read closely.
This matters most when a third party manages your property. In many of those arrangements, the management firm holds the contract for the property management system. You answer for the investment, yet you may have no direct relationship with the vendor holding the building's history and no admin login of your own. If you change management firms, that data relationship may not come with the building. And as agents take on more of the work, the logins they run on belong to whoever signed that contract.
Here's a way to find out where you stand. Ask what would happen if your largest platform vendor were bought next quarter by a company you'd never hire. Would the repair history, vendor performance records, and access logs come out in a form another tool could read? Could you stop the new owner from selling that data or training a product on it?
If you don't own your data & digital infrastructure, your vendors do.
Where the money is, and why you have to be able to prove it
Apollo's chief economist Torsten Slok has pointed to evidence that the return on AI spending outside Big Tech hasn't shown up yet. Treat any vendor's projected savings from agents as a claim to test.
If agents do deliver, the gain would show up in places you already track. Faster repairs help renewals. Less manual dispatching would lower the cost of running the building. Choosing vendors on their actual track record could lower what each work order costs. All of it flows into NOI, the income the property earns after operating expenses, and a buyer or lender values the asset on NOI divided by the cap rate, the yield they expect from it. At a 6 cap, every dollar of lasting NOI adds roughly sixteen dollars of value.
That value holds up only if you can prove the improvement. To show an agent cut your repair times, you need the repair times from before it went live. To show it lowered your cost per work order, you need that history in a form you control. If the only record lives inside the vendor's dashboard, the vendor is telling your lender the story.
As I said on PPP Podcast Ep. 50, "Optimism cannot become a substitute for evidence." Capture your baseline numbers in a system you control before the agent goes live. Then the before-and-after comparison is yours to show, and it comes with you if you move to a better tool.
Keep the data, choose the AI
Owners who can export and govern their own operating data get to choose which AI does the work. They can test one tool against another and drop the one that doesn't hold up. We help owners get there with a plan we call The PPP 5C™.
Clarify comes first, through a PPP Review. We map what operating history you have, where it lives, who holds the logins, and what your contracts let you export. Those answers are often scattered across contracts signed by different people at different times.
Connect and Collect would give you owner-controlled access to the systems in your buildings and gather their data into one consistent model. Work orders, vendor records, access logs, and meter readings would end up in a form you can reuse, whichever vendor produced them.
Coordinate would set the rules. You decide what an agent may read, what it may change, what needs a person's approval, and how long records are kept. Those permissions would belong to you and apply to every vendor.
Control is where the AI goes to work. Any agent, whether from your property management vendor, a specialist tool, or your own team, would act under your permissions on data you hold, so replacing a tool wouldn't cost you the history.
When one building's data is gathered in one place, held by you, and used under your rules, we call it the Property Brain™. It is vendor- and LLM-agnostic, meaning it works with any vendor's software and any AI model, so you can plug in the AI you choose and swap it later. Once that setup works on one property, you can extend it across your portfolio. That becomes the Portfolio Brain™, and it lets you compare repair times, vendor performance, and costs across every building you own.
Before you turn on the next agent
When a vendor offers agent features with a renewal or upgrade, ask for a complete export of one building's work order history first. See how long it takes, what format arrives, and whether there's a charge. That one request tells you more than the sales deck.
While you wait, confirm that you hold an admin login in your own name, separate from your management firm's. Then find the clause covering a sale of the vendor, a merger, or the end of the contract, along with any language about selling your data or using it to train AI. If that language is missing, get it written in before you sign.
Start with a PPP Review on one asset.
Own your data & digital infrastructure. Operate with strategic foresight. Build for the long game.
