Senior Housing & Care Didn’t Choose Fragmentation. It Inherited Two Systems of Record.

Part One: The Diligence File

Let me start where I actually started, which was not in software.

I spent more than two decades in commercial real estate before I ever wrote a word about data architecture. I bought buildings. I brokered them. I sat on both sides of the table — with operators trying to explain why their trailing twelve looked the way it did, and with investors trying to decide whether to believe them.

The pattern that eventually pushed me into building software was not fraud. In twenty-four years I have met very few people in this industry who were trying to deceive anyone. The pattern was worse than that, because it was structural.

Here is what a typical diligence process looks like. You request twelve months of operating data on a portfolio. What comes back is a census report exported from one system, a rent roll exported from another, a care-level revenue summary assembled by hand because no single system produces it, payroll registers from a third platform, agency invoices from a shared folder, incident logs from a fourth system, and a survey history reconstructed partly from memory.

Then the reconciliation begins. Census does not tie to the rent roll, because one counts units and the other counts residents, and neither agrees on how to treat a resident who moved to memory care on the fourteenth. Care revenue does not tie to the care plans, because the assessment that triggered the level change was documented in the clinical system nine days before anyone touched billing. Labor does not tie to acuity, because the scheduling system has no concept of acuity at all.

Six weeks later you have a number. Everyone signs. And the number is a reasonable estimate produced by intelligent people doing arithmetic on top of systems that were never designed to speak to each other.

Nobody lied. The numbers just came from too many places.

I did this often enough, on enough deals, that I stopped thinking of it as a diligence problem. It is not. Diligence is where the problem becomes visible, but the problem is present every single day in every single community, in the form of decisions made slightly too late on evidence assembled slightly too slowly.

Hold that six-week timeline in your head. In Part Eight I am going to tell you what the buyer on the other side of that table can now do in forty-five days, and why that gap is the most important number in this article.

Part Two: The Law of Mature Software Markets

A note on where this comes from. The framework in this section is not mine. The seven-stage lifecycle, the cross-industry comparison set, and the gravities framing below are drawn from Brendan Keeler’s “There Will Be Bundling” (Health API Guy, July 2025), which traces the pattern across healthcare, banking, automotive retail, property management, and several other verticals. If you have not read it, read it — it is the best single piece written on why systems of record behave the way they do. What follows here is an extension of his analysis into senior housing & care. The two-record collision is my contribution. The lifecycle it runs on is his.

Before I get to our industry, I want to establish something that is true well outside of it, because if you do not see the general pattern you will misdiagnose ours as a senior housing & care problem. It isn’t. It is a software-market problem that senior housing & care happens to be experiencing in an unusually severe form.

Every industry that digitizes ends up consolidated around a small number of systems of record, and the customers of those systems are unhappy and do not leave.

Consider three.

Banking. Roughly nine thousand banks and credit unions in the United States, and three companies — FIS, Fiserv, and Jack Henry — provide the core systems for the large majority of them. Survey after survey shows deep dissatisfaction with support, integration, and responsiveness. Switching runs in the low single digits per year against a much larger pool of contracts coming up for renewal. The reason is not loyalty. A core conversion is a multi-year, bet-the-institution project with no regulatory upside for attempting it.

Automotive retail. Two companies, CDK Global and Reynolds & Reynolds, account for roughly three quarters of franchise dealer management systems. In the mid-2010s both cut off third-party access to dealer data. Integration services priced in the tens of dollars became priced in the hundreds. Nearly a decade of litigation produced a settlement in the hundreds of millions — with no structural remedy, no behavioral restriction, and no admission of wrongdoing. The market emerged from a decade of antitrust conflict looking exactly as it did going in.

Multifamily real estate. A handful of property management systems dominate, and third-party integration means upfront fees, partnership review, and months of specialized engineering. This market has now had its own reckoning. The Department of Justice sued RealPage in August 2024 over algorithmic rent-setting; a proposed settlement was filed on November 24, 2025 with no fine, no admission of liability, a three-year court-appointed monitor, and a restriction limiting the algorithm to non-public data at least twelve months old. Separately, RealPage and thirty-seven property management companies settled a private class action for $359.9 million, with claims open into January 2027.

Three different industries. Three different regulatory regimes. Identical shape.

The Three Gravities

The best framework I have found for why this happens comes from David Yuan at Tidemark — whose framing I encountered through Keeler’s piece — describing a system of record’s moat as three compounding forces.

Workflow gravity. The system where the most users spend the most time. Once a workflow is habitual, moving it means retraining every person who performs it.

Data gravity. The system that creates and holds the most critical information and is the hardest to migrate. Not just current state — the history, the exceptions, the accumulated context.

Account gravity. The system whose sponsor is the highest-ranking person in the organization and which represents the largest financial commitment. Nobody gets fired for renewing.

Stack all three and switching stops being a decision. It becomes a hypothetical that gets raised in strategy sessions and never survives to a board vote.

The Seven Stages

Play this forward and every vertical runs the same lifecycle. These are Keeler’s seven stages, and I use his names for them. Innovation, where software arrives in a manual market and early entrants solve one painful problem deeply. Consolidation, where winners acquire rivals and the vendor field collapses. Platformization, where point solutions become suites and revenue expands through bundling, module count, and integration fees. Inertia, where customers stay because there is no viable exit and dissatisfaction and entrenchment rise together. Middleware rebellion, where integrators and standards bodies emerge to bridge legacy systems to modern tooling. Retaliation, where platforms respond with access throttling, partner restrictions, pricing changes, litigation, or acquisition-and-sunset. And finally resolution or stasis — either regulation forces openness, or the market hardens permanently.

Every vertical I named above is somewhere in stages five through seven. So is ours.

But ours has a wrinkle that none of them do, and it changes everything downstream.

Part Three: How Senior Housing & Care Got Two

Every mature vertical gets one system of record. One moat. One set of grievances.

Senior housing & care got two, from two different industries, and they arrived from opposite directions.

From Real Estate, Coming Down

Yardi® was founded in 1984 and remains privately held under its founder. It is, in every meaningful sense, a dominant system of record for institutional real estate, spanning multifamily, commercial, affordable, and senior housing.

Its senior housing & care suite is not a courtesy offering. Voyager Senior Housing sits at the core. Around it: the Care Suite for clinical documentation and medication administration, Senior CRM, the RentCafe Senior Housing & Care Portal for families, Senior IQ for business intelligence, and the Procure to Pay and Investment suites. That is a property management company that grew a clinical record — for an entirely rational reason. Senior housing is the one asset class where the operating business inside the building is as valuable as the building itself, and a PMS that cannot see care is blind to half the asset.

And in June 2026, Yardi® expanded Virtuoso Enterprise AI agents across leasing, accounting, maintenance, and month-end close. The platform now has an agent layer sitting on top of its own boundary.

From Healthcare, Coming Up

PointClickCare® was founded in 2000 by the Wessinger brothers, with Dave Wessinger now serving as CEO. As of its current 2026 materials, the company reports being trusted by more than 30,000 provider organizations, connecting roughly 2,800 hospitals and 3,600 ambulatory clinics, working with every major U.S. health plan, and supporting a marketplace of more than 400 integration partners. Its EHR is ONC-certified and spans skilled nursing, senior housing & care, and physician practice groups. Independent 2026 analyses estimate its share of the U.S. long-term care EHR market at roughly sixty percent.

The last publicly reported valuation was above $5 billion in 2022, on revenue then reported above $500 million growing around twenty percent annually. The company has remained privately held with Hellman & Friedman and Dragoneer among its investors, and has continued to expand — a next-generation EHR for practice groups, and an April 2026 strategic partnership with AIDA Healthcare targeting referral management and acute-to-post-acute transitions.

Outside of Epic, PointClickCare® is plausibly the most widely deployed electronic health record in America by facility count. MatrixCare®, owned by ResMed, and WellSky® are the credible alternatives at scale. Eldermark, ECP, ALIS, and August Health serve assisted living, where a private-equity roll-up dynamic is visible as well — LivTech assembling Eldermark and ChartPath, and adding home health through Alora in spring 2026.

That is a clinical company that grew census, billing, revenue cycle, and analytics. Also entirely rational: a chart that cannot see the ledger cannot tell an operator whether the care it documents is economically survivable.

The Stalemate

Both platforms expanded toward the same middle for the same reason every software company expands. It is always cheaper to sell another module to a customer you already have than to win a new one.

So the property management platform said: you already run your buildings on us — let us hold the chart.

And the electronic health record said: you already chart on us — let us hold the ledger.

Neither won. That is the entire story of senior housing & care technology, and everything else is commentary.

Because the fight did not resolve, the operator absorbed the cost. The typical multi-community operator is not running one system of record. They are running six to fifteen: property management, EHR, eMAR, CRM, payroll, scheduling and time capture, accounting, engagement, dining, maintenance, incident reporting, and an expanding set of sensors. Not one of those systems can see the business. The PMS knows the rent roll but not the acuity mix. The EHR knows the care plan but not the NOI. Payroll knows the hours but not whether those hours were clinically necessary.

Why This Gap Is Permanent

Here is the part I think most people in this industry get wrong, including some very smart people at both incumbents.

The gap is not a feature either vendor forgot to build. It is a consequence of where each of them came from.

A property management platform grew a chart. An electronic health record grew a general ledger. Neither grew the join — the connective tissue reconciling a resident, a day, a care event, an hour of labor, and a dollar across both worlds — because the join is the only thing that requires you to have come from both places simultaneously.

You cannot acquire your way to it, because the target has the same problem. You cannot build it inside one vendor’s boundary, because half the required data is definitionally outside that boundary. And you cannot standardize your way to it, because the two industries have no shared definitions and no shared regulator to impose any.

That is not a criticism of either company. It is a description of a structural condition. Structural conditions do not get patched in a release cycle.

Part Four: The Five Questions

Abstractions are easy to nod at and hard to act on. So let me make this concrete.

Here are five questions that determine whether a senior housing & care community succeeds or fails. Not reports. Questions. For each, what you need to answer it, and where that data currently lives.

1. Is our occupancy clinically appropriate and financially durable?

Occupancy is at a twenty-year high. NIC MAP reported senior housing occupancy across its 31 primary markets at 89.9% in the second quarter of 2026 — the twentieth consecutive quarter of increase — with occupied units reaching 639,650, up nearly 3,700 from the prior quarter. Fifteen of thirty-one primary markets were at or above ninety percent. Year-over-year inventory growth was 0.4%, with fewer than 16,000 units under construction nationally. Boston led at 93.3%; Miami trailed at 86.2%.

That is a spectacular demand environment. It is also the environment in which the most expensive mistakes get made, because when every unit fills, discipline about which resident fills which unit erodes.

To answer the question you need: current census by unit and care level, acuity distribution and trend, care plan changes over ninety days, length-of-stay expectation by acuity, referral source durability, and the rate structure attached to each.

Census lives in the PMS. Acuity lives in the EHR. Referral source lives in the CRM. Rate lives in billing. No system holds all four.

2. Can our staffing support the census we just sold?

Labor is the largest line in this business and it has not moved. Ziegler’s July 2026 CFO Hotline survey put employee compensation at 56.2% of operating budgets — essentially unchanged from 56.1% in 2025 — while ninety percent of organizations reported higher staffing costs. Turnover ran 35.4%, down slightly from 36.3%, with registered nurses at 41%, CNAs at 39%, and dining at 40%. CNAs were named the hardest role to fill by sixty-five percent of respondents. Staffing ratios reported around one caregiver per five to eight residents in memory care and one per six to ten in skilled nursing. Agency use remains common among multi-site operators, with more reporting frequent use in 2026 than the prior year even as many invest in reducing it.

To know whether staffing supports census you need: hours by role by shift, acuity-weighted demand, agency utilization and premium, overtime concentration, credential status and expiration, and callout patterns by community and shift.

Hours live in time and attendance. Acuity lives in the EHR. Agency spend lives in accounts payable, sometimes a spreadsheet. Credentials live in HR, sometimes a binder.

The question that actually matters — can this schedule support this census at this acuity — sits at the intersection of four systems, none of which was built to answer it.

3. Is the care we delivered the care we billed?

This is the quietest margin leak in senior housing & care, and in my experience the largest.

A resident’s needs increase. A caregiver notices. An assessment is completed. A care plan is updated. And then, somewhere between the clinical system and the billing system, the level-of-care change either propagates or it does not. When it does not, you deliver a higher level of care at a lower level of revenue, indefinitely, and nobody finds out until someone runs a variance analysis — if anyone ever does.

To catch it you need care plan change events, assessment history, billing effective dates, rate tables, and collections, joined at the resident level across a clinical system and a financial system that share no key.

Every operator I know believes this is happening somewhere in their portfolio. Almost none can tell me where.

4. What is our survey exposure right now?

A major regulatory shift in the last eight months made this materially harder rather than easier. Part Five is devoted to it.

5. What is this portfolio actually worth?

NIC MAP’s Senior Housing Market Outlook, Second Edition puts the current development pace at 10,445 units per year, projects a 2027 need above 141,000 units, and estimates approximately 1 million cumulative units of build-out by 2035. The outlook places the investment gap through 2050 at $1.0755 trillion. Its demographic baseline is equally direct: the population age 80 and older is projected to grow by approximately one-third by 2030 and nearly double by 2040.

Capital is arriving at scale to meet that. And it is underwriting operating performance on trailing spreadsheets and hand-assembled PDFs.

Notice the pattern. Every one of the five questions requires a join across the real estate record and the clinical record. Not one can be answered inside either. This is not five problems. It is one problem wearing five hats.

Part Five: The Deregulation Trap

On December 3, 2025, CMS issued an interim final rule — effective February 2, 2026 — repealing the 2024 minimum staffing standards for long-term care facilities. Gone are the 3.48 total nursing hours per resident day, the 0.55 RN hours, the 2.45 nurse aide hours, and the 24/7 registered nurse requirement. The statutory baseline reverts to eight consecutive hours of RN services daily and a full-time director of nursing.

The path there ran through two federal courts and Congress. The Northern District of Texas vacated the rule in April 2025; the Northern District of Iowa blocked portions that June. The reconciliation bill signed July 4, 2025 imposed a moratorium on implementation and enforcement running to September 30, 2034.

Most of the industry read that as relief. I understand why. I would encourage a more careful reading.

Because CMS did not repeal everything. The enhanced facility assessment requirements remain in effect, as do the Medicaid institutional payment transparency provisions.

Think carefully about what that combination means.

Under the mandate, staffing adequacy was a number. Someone in Washington set it, you either hit it or you did not, and your obligation was arithmetic. Burdensome, arguably unfunded — but legible.

Now the number is gone and the facility assessment remains. The obligation has shifted from meet this threshold to determine, document, and defend the staffing your resident population actually requires, using competency-based, acuity-informed evidence.

That is not a lighter burden. It is a heavier one wearing lighter clothing. A threshold you can satisfy with a payroll report. A judgment you must defend with a record.

And here is what should concern every operator reading this: you now carry the evidentiary burden for a determination you are making across systems that cannot see each other. Acuity is in the clinical system. Staffing is in payroll and scheduling. Census is in the PMS. The assessment itself is a document. When a surveyor asks you to demonstrate that your staffing plan reflects your resident acuity, you are assembling that answer manually, under time pressure, from four sources.

I have seen this movie before in commercial real estate. When a prescriptive rule is replaced by a principles-based standard, organizations with good records get relief and organizations with bad records get exposure. Deregulation is not neutral. It transfers risk from the regulator’s judgment to yours, and judgment requires evidence.

The staffing mandate is dead. The staffing data problem just became significantly more important.

Part Six: What Happens to Companies That Try to Close the Gap

If the join is the missing piece, the obvious question is why a third party has not simply built it by connecting to both sides. Several have tried, and we now have appellate case law describing what happens.

Real Time Medical Systems v. PointClickCare

Real Time Medical Systems is an analytics company serving skilled nursing facilities. For roughly a decade it extracted data from PointClickCare® on behalf of PointClickCare’s own customers, using automation, because that was the practical means available.

Then PointClickCare® entered the analytics business.

Real Time began encountering CAPTCHAs. Then harder ones. Then, on the record, indecipherable ones, followed by account lockouts.

Real Time sued in the District of Maryland in January 2024. Because the 21st Century Cures Act provides no private right of action, they used federal information blocking as the predicate wrongful act under Maryland unfair competition law — a construction that should get the attention of every system of record in every regulated industry. Judge Paula Xinis granted a preliminary injunction on July 29, 2024. The Fourth Circuit affirmed in a published fifty-eight-page opinion on March 12, 2025 (No. 24-1773), and denied rehearing en banc on April 23, 2025.

Read what the court actually found.

On the security defense: PointClickCare® “failed to articulate a specific security risk,” gesturing broadly at the general danger of bots. None of the security, performance, or manner exceptions applied.

On motive, the panel wrote that the timing of the introduction, escalation, de-escalation, and re-escalation of the CAPTCHAs — tracking the vendor’s own entry into the analytics market and its NDA-protected discussions with Real Time — was “highly suggestive” that the conduct was aimed at Real Time specifically.

On the public interest: it does not weigh in favor of such anticompetitive and harmful behavior.

And critically, the court held that once a plaintiff establishes facial information blocking, the burden shifts to the defendant to establish that an exception applies. That allocation matters more than any single fact in the case.

Current status, as of this writing: the district court granted a joint motion to stay proceedings on January 20, 2026, with a settlement conference held in April 2026. The parties have not litigated to a final merits judgment. What that means practically is that the Fourth Circuit’s published opinion stands as the operative precedent in this area, and is likely to remain so for some time.

What I Take From It

I do not cite this case to attack a competitor. Every company in this industry, including mine, will eventually be judged on how it behaves when a customer wants to leave.

I cite it because it is the clearest confirmation I have seen of something operators already feel: the data in your system is not being treated as your data. It is being treated as the vendor’s distribution advantage. And when a third party’s use of that data starts to look competitive, access becomes a lever rather than a service.

The Asymmetry Nobody Talks About

Now hold that case next to the RealPage settlement, and a striking asymmetry appears.

Information blocking regulation reaches developers of certified health IT. It does not reach a property management platform restricting access to rent roll, billing, sales pipeline, or financial data. There is no Cures Act for a rent roll.

Meanwhile the real estate side’s regulatory reckoning was about pricing coordination, not data access. It resolved in November 2025 with no fine, no admission, a three-year monitor, and a data-recency restriction on pricing models. Useful. Entirely orthogonal to whether an operator can get its own operating data out of its own system.

And the gap is widening rather than closing, because the clinical side keeps building infrastructure the real estate side has no equivalent of. TEFCA went live in December 2023 and had surpassed 500 million records exchanged as of February 2026, up from roughly ten million in January 2025, with FHIR alignment required of QHINs by July 2026. The CMS Health Tech Ecosystem, launched in July 2025, has drawn pledges from more than 700 organizations. Whatever you think of the pace, there is a national plumbing project for clinical data. There is no counterpart for operating and financial data in senior housing.

So the operator sits in the middle of an asymmetry:

• The clinical half of your record has federal protection, tested on appeal, with the burden shifted to the vendor, and a national exchange framework being built around it.
• The real estate and financial half of your record is protected by nothing except whatever your contract happens to say.

Most operators have never read that part of their contract. This is a good year to.

Part Seven: Scoring the Three Gravities Honestly

Apply Tidemark’s framework to our industry without flinching. I have done this many times, usually at two in the morning.

Workflow Gravity: The Incumbents Own It, and I Have No Interest in Fighting for It

The med tech at six in the morning is in the eMAR. The nurse is in the EHR. The leasing coordinator is in the CRM. The controller is in the accounting system.

Those workflows are not moving. Any company that tries to move them is signing up for a five-year replacement war against a vendor the state surveyor already recognizes, in an industry with 35.4% annual turnover where every retraining cycle is paid for twice.

This is why our position has always been that operators can keep the systems that work and that replacement is never required. That is not a limitation on the strategy. It is the strategy. Anyone telling you they will rip out your EHR, your PMS, and your payroll and replace all three is either inexperienced or selling you a decade of pain.

Data Gravity: Contested — and the Only Reason a Company Like Ours Can Exist

Here is the asymmetry that makes senior housing & care different from every vertical in Part Two.

In banking, the core has all the data. In auto retail, the DMS has all the data. In acute care, the EHR has all the data.

In senior housing & care, nobody has all the data.

The operating truth of a community is a join across two records that will never join themselves, held by two companies from two industries with no shared definitions and no shared regulator.

A layer that owns that join is not a dashboard. It is not a report. It is a record — and it is the only record from which the five questions in Part Four can be answered.

Let me be precise about the claim. I am not saying the incumbents lack data. They have enormous amounts of it; PointClickCare® accurately describes itself as holding the largest long-term and post-acute care dataset in existence. I am saying that the specific combination required to run this business is, by construction, split — and the entity that resolves that split holds something neither incumbent can replicate without becoming the other one.

Account Gravity: The Part Almost Everyone Gets Backwards

In every other vertical, account gravity belongs to the incumbent, because the system of record is the largest technology line item and its sponsor is the most senior person who signs technology contracts.

In senior housing & care, there is a buyer sitting entirely above the operator’s technology budget: the capital structure.

RIDEA and SHOP arrangements, where the owner participates in operating performance rather than collecting fixed rent, mean the owner’s return is a direct function of operating data quality. REIT compliance posture under §856 depends on operating structures that must be documented. Lender covenants are tested against operating metrics. GP reporting packages, quarterly asset management reviews, and diligence on the next acquisition all consume the same underlying record.

None of that was in the design brief for either an EHR or a PMS. The EHR was built for a nurse and a surveyor. The PMS was built for a property manager and a controller. Neither was built for an asset manager, a lender’s credit committee, or an investment committee memo.

I have spent my career on that side of the table. I can tell you exactly what it wants and has never had: one governed record it can trust without re-deriving it.

That is the opening. Not better reports. Not prettier dashboards. A different principal.

Part Eight: The Number That Should Worry Every Operator

I promised in Part One to come back to the six-week diligence file. Here is why it matters more in 2026 than it did in 2016.

The capital side has already built the data advantage. Operators have not.

Look at what the largest buyers are doing right now. Welltower closed 41 transactions totaling $3.2 billion in the first quarter of 2026, thirty-seven of them sourced off-market, then closed 30 transactions totaling $6.2 billion in the second quarter — again primarily off-market. Company leadership attributes this directly to a data science platform that has compressed its transaction timeline to forty-five to sixty days, letting it screen tens of billions of dollars of potential deals and act only on those meeting its criteria. Its same-store senior housing operating revenue grew 9.2% year over year in the second quarter, with RevPOR at $6,059 per month.

Ventas raised 2026 investment guidance to roughly $3 billion after approximately $6 billion of acquisitions since 2024, with SHOP now representing more than half of annual NOI. In the first quarter of 2026 alone it acquired 44 communities across fifteen states — about 5,100 units at an average of $325,000 per unit, at an average in-place occupancy of 88% — working with thirteen operators, four of them new relationships. Roughly two-thirds of those deals were sourced off-market. Ventas leadership, like Welltower’s, publicly credits its data and machine learning platform as the differentiator.

Now put the two sides of that table next to each other.

The buyer arrives with a machine-learning platform, a forty-five-day close, and a portfolio-wide view of comparable operating performance. The seller arrives with six weeks of manual reconciliation across nine systems.

That is not a fair negotiation, and it is not going to become fairer on its own. The information asymmetry in senior housing transactions has inverted in the last five years. It used to be that the operator knew the building and the buyer was guessing. Now the buyer has a model of your market, your comps, your rate trajectory, and your likely margin — and you have a data room.

I want to be careful here. I do not think this is predatory. Welltower and Ventas built exactly what any rational allocator would build, and the sector needs the capital they are deploying. But operators should understand the position they are negotiating from, and they should understand that the cure is not better negotiating. The cure is having a record as good as the one across the table.

This is also, incidentally, the clearest evidence for my whole thesis. The capital side did not solve its data problem by asking its software vendors to solve it. It built a governed record above its systems, because that is the only place the answer lives.

Part Nine: What an Operating Record Actually Is

The phrase “single operating record” has been so thoroughly abused in enterprise software that I hesitate to use it. Let me define what I mean precisely, including by saying what it is not.

It is not a dashboard. A dashboard is a rendering. If the underlying numbers are assembled differently each time, a dashboard just makes inconsistency arrive faster and in nicer colors.

It is not a data lake. A lake is storage. Copying six systems into one bucket without reconciling their semantics produces a large, expensive collection of disagreements.

It is not an integration platform. Pipes move data. They do not decide that this system’s “resident” and that system’s “resident” are the same person, or that a care level effective on the fourteenth means the same thing in both.

It is not an AI agent. An agent is a consumer of a record. It is not a record. That distinction is about to matter enormously.

An operating record has four properties. Miss any one and you do not have one.

1. Canonical entities. One resident. One community. One employee. One care event. Resolved across every source system, with a durable identity that survives a move, a readmission, a name change, and a system migration.

2. Governed semantics. A defined, versioned, documented meaning for every metric that matters. Not “occupancy” but this occupancy, computed this way, from these sources, as of this moment — with the definition itself under change control, so last quarter’s number and this quarter’s number came from the same rule, or you can see exactly when and why the rule changed.

3. Auditability. Every number traceable to source, with lineage and timestamp. In a regulated industry this is not a nice-to-have. It is the difference between an assertion and evidence.

4. Operator ownership. The record belongs to the operator, is exportable by the operator, and does not become someone else’s leverage. If the layer holding your operating truth can itself become a lock-in mechanism, you have not solved the problem. You have relocated it.

Here is the test I apply: Can the same record be defended, without reassembly, to a state surveyor on Tuesday, a lender’s credit committee on Wednesday, and a buyer’s quality-of-earnings team on Thursday?

If those three audiences require three different assembly processes, you do not have an operating record. You have three reporting projects and a hope.

Part Ten: Why AI Makes This Urgent Instead of Optional

I published Industry Findings in June documenting a thirty-day window. Between May 12 and June 18, 2026, PointClickCare® launched Advisor Suite, extended Chart Advisor into senior housing & care, seated AI-experienced board leadership, and signaled a company-wide AI strategy. Yardi® expanded Virtuoso Enterprise agents across leasing, accounting, maintenance, and month-end close on June 15.

I said then that I read those moves as validation and warning, and I meant both equally. Let me expand.

The Ceiling Is Architectural

An AI agent can only reason over the data it can see. That is not a limitation of the model. It is a property of the deployment.

An agent inside an electronic health record can reason brilliantly about clinical documentation and cannot tell you what a care plan change did to your margin. An agent inside a property management system can reason brilliantly about leasing velocity and cannot tell you whether the residents it just helped you move in are clinically supportable at your current staffing.

Both agents will be genuinely useful. Neither can answer any of the five questions in Part Four. And an operator who buys both now has two confident, articulate, well-designed systems producing partial answers with full conviction — arguably worse than having none, because partial answers delivered fluently are harder to distrust than no answer at all.

AI does not solve fragmentation. It industrializes it. Every silo now gets a spokesperson.

AI Also Closes Doors

There is a second-order effect that is underappreciated and that operators should plan for.

For fifteen years, open APIs were an unambiguous good for platform vendors. Integrations increased stickiness, expanded the ecosystem, and cost the platform almost nothing.

AI changed that math. When the value of a workflow can be captured by whoever holds the data and puts an agent on top of it, every open API stops looking like an ecosystem and starts looking like a leak. We have watched this play out across the broader software industry over the last two years — in messaging, project management, design tools, note-taking — platform after platform tightening data-use terms, adding rate limits, and routing distribution through their own marketplaces, always with reasonable-sounding justifications.

There is no reason to expect our industry to be exempt. If anything, expect it earlier, because the data here is more valuable and the customers are less able to leave.

And AI Raises the Evidentiary Bar

In a regulated setting, an agent that cannot show its work is not an asset. It is a liability.

If an AI system recommends a staffing level and a surveyor asks what that recommendation was based on, “the model suggested it” is not an answer — particularly now that the facility assessment obligation has shifted the burden of proof onto you. If an agent flags a revenue integrity issue and the auditor asks for lineage, a confident paragraph of prose is not evidence.

This is why grounding, provenance, and audit trails are not compliance overhead bolted onto an AI strategy. They are the AI strategy. An agent reasoning over a governed record with traceable lineage is defensible. An agent reasoning over an unreconciled pile is a well-spoken guess.

The field has moved. Agents are no longer the differentiator. Where the agents live, which record they reason against, and who owns that record — that is the differentiator now.

Part Eleven: What Operators Should Do Monday, Regardless of What They Buy

This is the section I would most like people to use, and it is deliberately vendor-neutral. Everything below is worth doing whether you work with us, with a competitor, or with nobody.

1. Read your data rights clauses. All of them. Every contract says the data is yours. Find out what that means operationally. Can you export it? In what format? At what frequency? At what cost? Within what timeframe on termination? Does the agreement distinguish between your data and “derived” or “aggregated” data — and if so, who owns the derived layer? That last question is the one that matters most in an AI era, and it is the one almost nobody negotiated.
2. Test the export before you need it. Ask for a full extract this quarter. Not a report — an extract, with schema. If it takes six weeks and a professional services engagement, you have learned something important while it is still cheap to learn.
3. Negotiate exportability at renewal, not at exit. The only moment you have leverage over a system of record is the moment before you sign. Machine-readable export, defined format, defined SLA, defined cost, no charge on termination, transition assistance period. Put it in writing while someone wants your signature.
4. Write down your metric definitions. Not the software’s definitions. Yours. How does your organization define occupancy, acuity tier, agency reliance, labor as a percentage of revenue, same-store variance? Get them on paper, approved, versioned. You will discover in the process that two departments have used the same word differently for years. That discovery alone is worth the exercise.
5. Instrument the care-to-billing gap specifically. Pick one community. Pull ninety days of care plan changes. Match them to billing effective dates. Measure the lag and the leakage. Extrapolate across the portfolio and decide whether the number is acceptable.
6. Rebuild your facility assessment as a living artifact, not an annual document. The mandate is gone as of February 2, 2026; the assessment obligation is not. It should tie to actual acuity data, refresh on a defined cadence, and carry its evidence with it. If your assessment is a Word document updated once a year, you are carrying unpriced risk.
7. Ask every new vendor the exit question first. Before features, before pricing: what does leaving look like? A vendor who answers that clearly and in writing is telling you something important about how they intend to compete.
8. Give your capital partners one record instead of one package per request. If your lender, your REIT partner, and your board each get a bespoke assembly, you are paying for the same work three times and creating three chances to disagree with yourself.
9. Prepare your record before you are in a process, not during one. Given what Part Eight describes about how fast the buy side now moves, the operator who can produce clean, lineage-backed operating data in days rather than weeks is negotiating from a materially different position.
10. Decide who owns the operating record inside your organization. Not the EHR. Not the PMS. The record. If nobody owns it, it does not exist, and every question in Part Four will keep taking six weeks.

None of that requires buying anything. All of it improves your position regardless of what you eventually buy.

Part Twelve: The Four Moves Coming, and How to Read Them

These four moves are what Keeler’s retaliation stage looks like when it arrives in our vertical. I would rather name what I expect in public than pretend it is not coming.

Move one: native equivalents, marketed as “already integrated.” Already happening, and it costs the incumbents nothing. What it means for you: an agent inside one vendor’s boundary is genuinely useful within that boundary. Buy it for what it does. Do not mistake it for an answer to the cross-system question, and ask specifically which of your other systems it can see.

Move two: access terms tighten. Rate limits, partnership gates, certification requirements, revised language about how data obtained through the platform may be stored, used, and shared. What it means for you: this is why item three above is time-sensitive. Terms are easiest to negotiate before they are tightened.

Move three: bundling pressure. Whatever an independent layer charges for becomes free-with-suite. What it means for you: free-with-suite is a real value proposition and deserves honest evaluation. Just evaluate it on whether it answers the question, not on whether it costs nothing.

Move four: acquisition. If the operator-controlled record becomes something capital partners request by name, someone will try to buy it and quietly retire it. As Alex Johnson of Fintech Takes puts it, the cores are where good tech in banking goes to die — a line I first read in Keeler’s piece. What it means for you: ask any vendor, including us, what happens to your data and your terms in a change of control. Get it in the contract.

Part Thirteen: What Has to Be True

I am not going to spend nine thousand words arguing that structure beats product and then exempt my own company. Here are the conditions under which I am wrong.

The join has to be genuinely hard. If reconciling a canonical resident, a shared acuity definition, and an auditable trail across a PMS and three clinical systems is a mapping exercise rather than a semantics problem, the incumbents replicate it in two quarters and this thesis dies. I believe the difficulty is definitional rather than technical — but that belief has to survive contact with real portfolios. It is precisely why our beachhead is operators running ten to forty communities rather than a logo hunt. That segment is large enough to have the problem in full and small enough that you cannot hide behind a corporate analytics team.

Operators have to assert ownership in practice, not just in principle. Every contract says the data belongs to the customer. Real Time is what that principle looks like under pressure. If operators do not start negotiating data access the way they negotiate rate and term, no architecture will save them.

The capital side has to demand the record from operators, not just build it for itself. This is the condition I have the most conviction about and the most bias toward, so weigh it accordingly. Welltower and Ventas have demonstrated that a governed operating record is worth building — they built one. What has not yet happened is that requirement flowing downstream into term sheets and management agreements. If the demand environment stays forgiving enough that precision does not pay, account gravity does not shift and my opening narrows considerably. I think a sector that faces a current development pace of 10,445 units per year against a projected 2027 need above 141,000 units will not stay forgiving. But I could be early, and early is indistinguishable from wrong for longer than most companies survive.

The AI framing has to be temporary. “Agents grounded in a governed record” is the correct architecture argument in 2026. It will be a commodity argument by 2028. Anything durable comes from owning the join and serving the capital side — not from the agents. Anyone selling you agents as the differentiator, including me on a bad day, is selling you last year’s insight.

And the industry has to stay unconsolidated enough to matter. If a single transaction genuinely merged a dominant PMS and a dominant EHR into one book of record, the split would close and my central claim would weaken considerably. I do not expect it — the cultural, regulatory, and technical integration is brutal, and both sides have tried the organic version and produced modules rather than joins. But it is the scenario I watch for.

Part Fourteen: A Five-Year View

For whatever it is worth, here is where I think this goes.

2026–2027. AI agents ship inside every major platform and become table stakes. Operators buy several. Satisfaction is high on narrow tasks and low on cross-system questions. “Our AI is already integrated” enters every sales deck in the industry. Occupancy crosses ninety percent nationally, which NIC has already signaled is likely this year, and the resulting margin conversation exposes how few operators can attribute their own performance.

2027–2028. Data access terms tighten industry-wide, justified on security and performance grounds. At least one more information-blocking-style dispute reaches a published decision in post-acute care. Operators discover their export rights are thinner than assumed. Data rights become a standard diligence item in senior housing transactions — I would argue they should already be one.

2028–2029. Capital partners begin requiring standardized operating data as a condition of financing rather than a courtesy of reporting. This is the inflection point. When a lender’s term sheet or a management agreement specifies the record, the record stops being an IT decision and becomes a covenant.

2029–2031. Either a standard emerges — most likely driven by capital and trade associations rather than vendors, because vendors have no incentive to standardize what they monetize — or the market hardens into stasis and the cost of fragmentation gets permanently capitalized into cap rates. I do not know which. I know the deciding factor is whether operators and their capital partners act collectively in the next thirty-six months or keep solving it one portfolio at a time.

Why This Matters More Than the Software

I got into senior housing because of what it is, not because of what it runs on.

We are talking about where people live at the most vulnerable stage of their lives, staffed by people who work harder for less recognition than almost anyone in this economy. A CNA in a memory care community makes more consequential decisions before nine in the morning than most executives make in a quarter. And in 2026, sixty-five percent of operators say that role is the hardest one they have to fill.

That caregiver is documenting the same resident in three systems because two software companies from two different industries each decided, reasonably, to grow toward the middle and neither one won. That duplicate documentation is not an abstraction. It is a real cost, paid in minutes that should have gone to a resident, multiplied by every shift, in every community, every day, for twenty years.

We did not choose that. We inherited it.

But inheriting a structure is not the same as being stuck with it. This industry faces an approximately 1 million-unit cumulative build-out requirement by 2035 with a workforce turning over at 35% a year, under a regulatory regime that just moved the burden of proof onto operators, negotiating with capital partners who can now close a transaction in forty-five days using models we cannot see.

The operators who take control of their own operating record over the next thirty-six months will be the ones who can answer a lender’s question in an afternoon, defend a staffing decision to a surveyor with evidence instead of narrative, absorb whatever technology arrives next without another integration project, and keep their people at the bedside instead of at a keyboard.

That is the work. Everything else is a module.

John Hauber is Founder and CEO of SeniorCRE and leads HavenCo, Haven Senior Investments, Haven Senior Housing & Care Partners, and Haven Realty. He has spent more than twenty-four years in commercial real estate and senior housing investment.

Sources

• Brendan Keeler, “There Will Be Bundling”, Health API Guy, July 9, 2025 — source of the seven-stage systems-of-record lifecycle, the cross-industry comparison set used in Part Two, and the Alex Johnson quotation in Part Twelve.
• NIC MAP, Q2 2026 senior housing data release (July 9, 2026); NIC MAP Q1 2026 release; and Senior Housing Market Outlook, Second Edition (year-end 2025 data, published in 2026).
• Ziegler CFO Hotline, 2026 State of the Senior Housing & Care & Care Workforce (July 2026).
• CMS, Medicare and Medicaid Programs; Repeal of Minimum Staffing Standards for Long-Term Care Facilities, interim final rule (90 FR, December 3, 2025; effective February 2, 2026); Public Law 119-21 (July 4, 2025) enforcement moratorium through September 30, 2034; American Health Care Association v. Kennedy (N.D. Tex. 2025); Kansas v. HHS (N.D. Iowa 2025).
• Real Time Medical Systems, Inc. v. PointClickCare Technologies, Inc., No. 24-1773 (4th Cir. March 12, 2025); en banc denied April 23, 2025; D. Md. No. 8:24-cv-00313 docket (stay granted January 20, 2026).
• United States v. RealPage, Inc., proposed final judgment (November 24, 2025); related private class settlement of $359.9 million.
• PointClickCare® corporate materials and press releases, 2025–2026; Hellman & Friedman and Dragoneer investment announcements; third-party LTC EHR market analyses, 2026.
• Yardi® product and press materials, 2025–2026, including Yardi Virtuoso Enterprise announcements (June 15, 2026).
• Welltower Q1 and Q2 2026 earnings disclosures and coverage; Ventas Q4 2025 and Q1 2026 earnings disclosures and 2026 investment guidance.
• ASTP/ONC and CMS interoperability materials: TEFCA volume as of February 2026; CMS Health Tech Ecosystem (launched July 2025).
• Tidemark (David Yuan), Vertical SaaS Knowledge Project, on workflow, data, and account gravity — encountered via Keeler, above.

Yardi®, PointClickCare®, MatrixCare®, and WellSky® are trademarks of their respective owners and are referenced here nominatively. Product characterizations reflect publicly available product materials as of August 1, 2026 and are not advertised-capability claims.