Every CEO in property is being told the same thing right now: adopt AI or fall behind. It is framed as a source of advantage, the way ERP was once framed, or the internet, or mobile. There is a problem with that framing, and it is worth stating plainly before you spend hard against it. Your competitors are buying the same AI you are, from the same short list of providers, at close to the same price, through the same interfaces. If the advantage is the engine, and everyone can rent an identical engine by the hour, then it is not an advantage. It is a cost of doing business.

That is not an argument against AI. It is an argument about where the advantage actually sits, because the answer changes what a CEO should be spending time and capital on.

The commodity is arriving faster than the strategy decks admit

Sit with how quickly the model layer has converged. Stanford's AI Index, which is about the most sober scorekeeper the field has, tracked the gap between the best proprietary model and the best open one at roughly eight percentage points in early 2024. A year later that lead had narrowed to under two. By early 2026 the top of the market was a crowded pack: six providers clustered within a narrow band of scores, and the Index's own read was that competitive pressure had moved off raw model performance and onto cost, reliability, and domain-specific fit.

When six vendors are separated by something close to a rounding error, "we have the best model" stops being a strategy. This is what commoditization looks like in real time. The capability becomes a utility, available to everyone on roughly equal terms, and a capability available to everyone on equal terms cannot, by definition, be your edge. The strategist's instinct here is old and correct: you do not build a moat out of an input your competitor can order from the same catalog.

So if the model is not the advantage, the honest question for a CEO is what is, and whether the company is actually building it.

The case that your data is the moat

The strongest answer, and the one most of the serious analysis has converged on, is that as the model commoditizes, the scarcity moves to the data. Andreessen Horowitz made this point directly: when foundation model capability becomes a shared resource, the durable advantage shifts to proprietary data that rivals cannot buy, borrow, or scrape. The model is trained on the public internet. Your edge is precisely what is not on the public internet.

For a property company, that is a real and specific asset. The accumulated record of how your portfolio actually behaves: your leasing histories, your renewal and delinquency patterns, your maintenance and turn data, the economics of your specific owner relationships, the operational signals your buildings throw off every day. No competitor has that, and no general model can infer it, because it was never public. Point a commodity intelligence at a proprietary record like that and it can produce things a rival running the identical model cannot reproduce, because the rival does not have your history to feed it. That is what a moat in the AI era looks like: not a better engine, but better fuel that only you possess.

It is a clean story. It is also not quite true yet for most companies, and a CEO should hear the hard part before believing the easy part.

The case against, which most CEOs need to hear

Having data is not the same as having a moat. Bessemer Venture Partners has made the skeptical case worth taking seriously: modern models already know an enormous amount from public data, and basic AI features are becoming free table stakes bundled into everyday software, so simply owning a pile of records buys you less differentiation than the "data is the new oil" pitch implies. A generic model with a little domain tuning can rival one trained on a large private corpus more often than data-moat evangelists admit.

The deeper problem is closer to home. For most property operators, their data is not an asset at all. It is a liability. It sits fragmented across a property system, a separate accounting system, a bank feed, and a dozen spreadsheets, none of which agree with the others. It is inconsistent, undocumented, and frequently locked inside a vendor platform the company does not control. A commodity model pointed at that produces confident nonsense, faster. A pile of disconnected, contradictory records is not a moat. It is just storage you are paying for. So the moat thesis comes with a precondition, and it is a precondition most CEOs have quietly not met.

What actually resolves this

The resolution is not "AI wins" and it is not "data wins." It is narrower and more useful than either. The advantage belongs to the company whose proprietary operational data is structured, connected, and under its own control, so that when the commodity intelligence is applied, there is something coherent and exclusive underneath it. Each of those three words is doing work. Structured, or the model cannot use it. Connected, or it contradicts itself. Under your control, or it is not yours to compound.

That last condition is the one CEOs miss most, and it is the most strategic. If your operational history lives inside a point solution you licensed rather than a system you own, the vendor holds the moat, not you. Your leasing and maintenance records become an asset on someone else's balance sheet, and you rent access to your own advantage, for as long as the relationship lasts and on terms you do not set. This is why the system-of-record question is a CEO decision and not something to delegate to IT as a procurement exercise. It determines whether the company is accumulating an owned, compounding asset or renting one that walks out the door with the vendor.

This is the case for keeping the operational and financial record in one system the company actually controls. RIOO is built directly on NetSuite for that reason: when the work order, the ledger entry, and the owner record are one record inside a system you own, the data that could become an advantage compounds in a place a competitor cannot copy and a departing vendor cannot take with them.

The advantage was never the AI

The CEOs who win the AI era in property will not be the ones who adopted AI first. Everyone adopts the same AI, and that race ends in a tie. They will be the ones who spent the unglamorous years turning their operational history into a structured, connected, owned asset, so that when the commodity intelligence arrives, and it is arriving quickly, they have something to point it at that no competitor can reproduce.

Adopt the AI. You have to; it is table stakes now, and refusing it is a disadvantage even though accepting it is not an advantage. But do not confuse buying the commodity with building the moat. The advantage was never the AI. It was always what you feed it, and whether you own what you feed it.

FAQs

Q1. If AI is a commodity, should we still invest in it?
Yes. Commodity does not mean worthless, it means non-differentiating. Electricity is a commodity and you would not run a business without it. Not adopting AI is a real disadvantage even though adopting it is not, on its own, an advantage. The mistake is budgeting for it as a source of edge rather than as a cost of staying competitive, and then neglecting the thing that actually differentiates.

Q2. Isn't proprietary data automatically an advantage?
No, and this is the most expensive assumption in the category. Data becomes an advantage only when it is structured enough for a model to use, connected enough to be internally consistent, and owned enough to compound under your control. Fragmented, contradictory, or vendor-locked data is a liability wearing the costume of an asset.

Q3. What data actually counts as a moat for a property company?
The longitudinal operational record a competitor cannot obtain from the public internet: leasing and renewal histories, delinquency and turn patterns, maintenance behavior across your specific assets, the economics of your owner relationships. It is valuable precisely because it is specific to your portfolio and accumulated over years, which is what makes it hard to replicate.

Q4. Why does it matter who owns the data if it is technically ours?
Because control and portability are what make an asset yours in any way that counts. If the record lives inside a platform you licensed, the vendor sets the terms of access and export, and your advantage is contingent on that relationship. When the vendor changes or you leave, an asset you thought you owned can prove surprisingly difficult to take with you.

Q5. Is this a technology decision or a strategy decision?
Strategy, which is why it belongs on the CEO's desk. The question is not which product has which feature. It is whether the company is building a compounding, owned data asset or renting a fragmented one. That is a decision about durable competitive position, and it happens to express itself through a systems choice.

Q6. How long does a data moat take to build?
Years, and that is the point. An advantage you can assemble in a quarter is one your competitor can assemble in a quarter too. The defensibility comes from accumulation and refinement over time, which cannot be bought overnight at any price. The implication is that the cost of starting late compounds, so the decision is more urgent than it feels.

Q7. Our competitors manage more units than we do. Are we already behind on data?
Not necessarily. Raw volume matters less than whether the data is structured, connected, and owned. A disciplined operator with a clean, consolidated record can extract more usable advantage than a larger rival whose data is scattered across incompatible systems. More units with worse data is not a lead you should envy.

Q8. Where should a CEO actually start?
Not with an AI vendor. Start with an honest look at whether your operational and financial record is consolidated and under your control, because that is the precondition everything else depends on. It is the least exciting item on the agenda and the one that determines whether any of the AI spending downstream produces an advantage or just parity.

Q8. Does this mean we should delay adopting AI until the data is ready?
No. Run both tracks. Use the commodity layer where it helps operations today, and separately do the slower work of turning your operational history into an owned asset. Just keep the two straight in your own mind: the first keeps you level with the field, and only the second can put you ahead of it.