For most of its history, the ERP had a humble and slightly thankless job. It was where the business wrote things down. A lease was signed, so it went into the system. Rent was collected, an invoice was paid, a work order was closed - each event, after it happened, was recorded, validated, and stored so that later, someone could trust the number. The ERP was the organization's memory. Reliable, authoritative, and fundamentally backward-looking. It told you, with great precision, what had already occurred.

That job is not going away. But it is being quietly demoted from the whole point of the system to merely its foundation, because something is being built on top of it that changes what the ERP is for. The same platform that spent forty years recording the past is now being asked to interpret the present and act on the future - to notice that a building is trending toward a bad quarter, to propose the renewal price, to flag the vendor invoice that does not match the contract, and increasingly, to take the routine action itself. The system of record is becoming a system of intelligence.

This is not a feature release. It is a change of category, and it has consequences that reach well past the software. When a system's job was to remember, the stakes of a gap in its data were bounded - a missing record was an inconvenience you worked around. When a system's job becomes to decide and act, the same gap becomes a wrong decision executed automatically, at machine speed. The promotion of the ERP from record-keeper to decision-maker raises the stakes on the data underneath it by an order of magnitude, and most property companies have not yet noticed that the ground has moved.

Three Eras of Enterprise Software, and the One We Just Entered

To see why this matters, it helps to place it in a longer arc. Enterprise software has moved through recognizable eras, each defined by what the software was fundamentally for.

The first was the system of record. Its purpose was to capture and hold authoritative data - the general ledger, the lease abstract, the tenant roster. Industry strategist Geoffrey Moore gave these categories their durable names, and the system of record was the foundation: the trusted account of what is true. In property, this is the layer that made sure the rent roll and the balance sheet agreed.

The second was the system of engagement. As software moved to the cloud and onto phones, the emphasis shifted from storing data to putting it in front of people at the moment they needed it - portals, mobile apps, dashboards, the tenant self-service experience. Engagement made the record accessible and interactive, but it did not change the underlying logic. A human still read the information and decided what to do.

The third era is the one now arriving, and it was named before it was fully possible. In 2017, Greylock investor Jerry Chen argued in an influential essay on the next defensible business model that the coming moat in enterprise software would be the system of intelligence - an application layer that sits on top of the record, combines proprietary data across sources, and generates insight and action that competitors cannot easily replicate. When he revisited the thesis in 2023, after generative AI arrived, his update was sharp: foundation models were becoming table stakes, available to everyone, and so the durable advantage had moved to the proprietary data a system of intelligence sits on and what it does with that data.

That is the transition compressing into the ERP right now. The ERP is uniquely positioned to become the operational foundation of the enterprise's system of intelligence for one simple reason: intelligence is only as good as the record it reasons over, and the ERP already holds that record. In some organizations the reasoning layer will extend beyond the ERP itself - into a separate orchestration layer, an agent platform, or connected models - and that is a legitimate architecture. But wherever the reasoning lives, it is powerless without an authoritative operational record beneath it, and the ERP is the one system in the building that holds that record. The system that remembers is becoming the operational bedrock of the system that reasons, because it is the one system with the memory to reason with.

What "System of Intelligence" Actually Means for an ERP

It is easy to let a phrase like this dissolve into marketing, so it is worth being concrete about what changes. A system of intelligence differs from a system of record along three specific dimensions, and each one is visible in what the newest ERPs are shipping.

It reasons rather than merely retrieves. A system of record answers the question you ask it: what was occupancy last month, what is the balance on this account. A system of intelligence answers questions you did not think to ask: which of your buildings is quietly trending toward a bad year, which lease renewals are at the highest risk of churning, which maintenance pattern signals a real problem rather than bad luck. The data was always there; the older system just never interpreted it.

It acts rather than merely informs. This is the sharpest break. A dashboard tells a person that an invoice looks wrong and waits for them to do something. An agentic system flags the discrepancy, applies the rule, and - within administrator-defined policies - resolves it, escalating to a human only for the exceptions. The enterprise-software vendors have made this shift explicit. Oracle NetSuite now embeds agentic workers that autonomously execute multi-step processes such as collections follow-ups and expense handling within administrator-defined boundaries; SAP has framed its direction as the "autonomous enterprise"; and Forrester now describes the ERP itself as evolving into an intelligent orchestration engine rather than a passive database. Gartner has increasingly described AI-enabled ERP capabilities as becoming a baseline expectation rather than a competitive differentiator - which is the surest sign that a change has stopped being optional.

It is judged by outcomes rather than accuracy alone. A system of record has one core obligation: be correct about the past. A system of intelligence inherits that obligation and adds a harder one: be useful about the future without being reckless about it. That is a genuinely different engineering and governance problem, and it is why the same vendors racing to add agents are simultaneously adding audit trails, ROI dashboards, and approval boundaries. Intelligence without governance is not an asset; it is a liability that acts on its own.

Why This Lands Hardest in Property

Every industry is feeling this shift, but property feels it with unusual force, because property is an operations-intensive business where nearly every meaningful decision is a cross-functional read of the same core entities: properties, units, leases, residents, work orders, and money. A renewal price is a leasing decision, a revenue fact, an occupancy assumption, and a resident-relationship judgment at the same time. That is precisely the kind of decision a system of intelligence is built to make - and precisely the kind that falls apart if the underlying data is scattered or contradictory.

This is where the promotion of the ERP collides with how most property companies are actually built. The typical operator does not run one system; it runs a loose federation of them - a leasing tool, a maintenance app, a screening service, an accounting package - each holding its own slice of the truth. For a system of record, that fragmentation was a tolerable inefficiency; a human analyst reconciled the differences before anything important happened. For a system of intelligence, it is disqualifying. An agent asked to act on data that three systems define differently will either stall, which destroys the value of automation, or pick one version and act, which turns a quiet data discrepancy into a wrong action taken at speed.

There is a second reason property is exposed, and it is more hopeful. Property generates unusually rich, long-lived operational data - every unit has a maintenance life, every tenant a payment and behavior history, every building years of cost and occupancy patterns. That is the raw material a system of intelligence runs on. Property companies are, in this sense, sitting on better fuel than most industries. The question is whether it has been captured in a form the intelligence can actually use or not. The intelligence layer sets the ambition; the depth and continuity of the record set the ceiling.

The Data Is the Asset the Intelligence Runs On

Here is the part that should reframe how a property leader thinks about their ERP investment, because it inverts the usual assumption.

The instinct is to treat the intelligence as the valuable thing and the data as the plumbing that feeds it. In a world where every operator can buy access to broadly similar AI, that instinct is exactly backwards. The models are becoming a commodity - available to your competitor on the same terms and roughly the same timeline as to you. What is not a commodity, and cannot be bought at any price, is your proprietary operational history: the specific, accumulated, well-structured record of how your portfolio has actually behaved over years. That record is the asset. The intelligence is just the engine that finally converts it into advantage.

This is the deeper meaning of Jerry Chen's point that foundation models are table stakes. When intelligence is universal, the moat moves to the data the intelligence sits on. Two property companies can deploy the identical agentic ERP and get entirely different value from it, because one has fed it a deep, continuous, unified record and the other has fed it four disconnected fragments with the interesting history missing. The system of intelligence does not create the advantage. It reveals whether you spent the previous years building the asset or discarding it.

Which leads to a conclusion that sounds paradoxical but is not: the most important decision in adopting an intelligent ERP is not which intelligence you buy. It is whether your operational data is unified, deep, and owned enough to be worth reasoning over in the first place. The intelligence is the visible, exciting part. The data foundation is the part that actually determines the outcome - and it is the part that takes years, not a purchase order, to get right.

What This Asks of a Property Leader Now

If the ERP is being promoted from the system that remembers to the system that decides, then the work in front of a property company is less about chasing the newest AI feature and more about making sure the foundation is ready to be reasoned over and acted on. A few plain priorities follow from everything above.

Treat the unification of your operational data as the prerequisite it now is, rather than a project you will get to eventually. An intelligence layer on top of a fragmented record does not produce intelligence; it produces confident mistakes. The single-record foundation platform - where leasing, finance, maintenance, and resident management operate on one shared data model and dashboards and reports read directly from where the work happens - is precisely the substrate a system of intelligence requires. The order matters: unify first, then let the intelligence act.

Take the depth and continuity of your history seriously as an asset, because it is the one input to the intelligent ERP that money cannot shortcut. And insist on governance as a first-class requirement, not an afterthought - an ERP that acts on your behalf needs auditability, clear boundaries, and human escalation for exceptions, or its intelligence becomes a risk rather than an advantage.

None of this is glamorous, and that is exactly why it will separate the operators who convert the intelligent ERP into results from the ones who buy it and wonder why it underperforms. The intelligence will be broadly available. The readiness to use it will not.

The Quiet Reframe

For forty years, the question a property company asked of its ERP was a question about the past: what happened, and can I trust the number? The system of record answered it well, and that answer still matters. But the question is changing, and the new one is harder: what is about to happen, what should we do about it, and how much of that can the system handle on its own?

An ERP that can answer the second question is no longer a filing cabinet with good search. It is the operational foundation of a system of intelligence - the layer where the business increasingly decides and acts. The property companies that understand this early will spend the coming years doing the unglamorous work of unifying and deepening the record so that when the intelligence arrives, it has something worth reasoning over. The ones that don't will buy the same intelligence, point it at a fragmented history, and discover that a system of intelligence is only ever as intelligent as the foundation it was promoted from.

Frequently Asked Questions

Q1. What does "the ERP is becoming a system of intelligence" mean?
It means the ERP is shifting from a system whose main job was to record and store authoritative data about the past (a system of record) to the operational foundation of a layer that interprets the present and acts on the future - surfacing insights, recommending decisions, and, through agentic AI, autonomously executing routine workflows within administrator-defined policies. The record-keeping function remains, but the primary value increasingly comes from the intelligence and action layered on top of it.

Q2. Where does the term "system of intelligence" come from?
It was popularized by Greylock investor Jerry Chen in his 2017 essay "The New Moats," which argued that AI-powered applications sitting on proprietary data would become the most defensible layer in enterprise software. It builds on the older lineage of "systems of record" and "systems of engagement" associated with strategist Geoffrey Moore. Chen's 2023 update added that, once foundation models became widely accessible, the durable advantage shifted to the proprietary data a system of intelligence reasons over.

Q3. How is a system of intelligence different from a system of record?
A system of record retrieves and protects authoritative data and is judged on accuracy about the past. A system of intelligence reasons over that data to answer questions no one explicitly asked, acts on its conclusions rather than only informing a human, and is judged on the usefulness and safety of its outcomes. The record is backward-looking and passive; the intelligence layer is forward-looking and, increasingly, active - though it depends entirely on the record beneath it.

Q4. Does this mean property companies need to replace their ERP?
Not necessarily. Much of the shift is arriving as agentic and AI capabilities embedded into or layered onto existing cloud ERPs rather than requiring a rip-and-replace. The reasoning layer may even live partly outside the ERP. The more important readiness question is not which AI you adopt but whether your operational data is unified, deep, and owned enough to be reasoned over safely. An intelligence layer on top of fragmented data produces confident errors, not intelligence.

Q5. Why does data unification matter more once the ERP becomes intelligent?
Because the stakes of fragmentation change. When systems disagreed and a human read the data, the human reconciled the discrepancy before acting. When an autonomous agent reads data that multiple systems define differently, it either stalls or acts on the wrong version at machine speed, converting a quiet data conflict into a wrong action. Unified operational data is what lets an intelligent ERP act reliably rather than dangerously.

Q6. What should a property company do first?
Prioritize the foundation over the feature. Unify operational data onto a single shared record so leasing, finance, and maintenance cannot disagree; protect the depth and continuity of your historical data as a strategic asset; and require governance - auditability, boundaries, and human escalation - before allowing an ERP to act autonomously. The intelligence will be broadly available; readiness to use it well is the real differentiator.