The 70% Problem: Why Senior Living CFOs Are Overspending on AI and Under-Investing in the Boring Stuff 

It usually falls apart in a board meeting. Finance puts up a beautifully AI-generated forecast, all confident lines and clean projections, and then someone across the table gently asks why it doesn’t match the occupancy figure operations shared last Tuesday. A pause. A shuffle of papers. And the quiet, dawning realisation that the two systems have never once agreed on what a “resident day” actually means. The tool wasn’t wrong, exactly. It just answered a question nobody had bothered to define properly. 

If that scene feels a touch too familiar, you’re in good company. 

That gap between the dazzling demo and the disappointing Tuesday isn’t a you problem. It’s practically the industry’s default setting. 

The numbers are blunt about it. A landmark MIT study found that a staggering 95% of enterprise generative AI initiatives delivered no measurable business return, despite some $30 to $40 billion being poured in. Ninety-five! And before anyone blames the algorithms, the research is rather clear that the models were rarely the culprit. 

So what is? 

Well, this is where it gets interesting. BCG put a number on it that ought to be pinned above every finance director’s desk: only about 10% of AI success comes from the model itself, another 20% from the platform, and a thumping 70% from the boring stuff. The data. The processes. The people. 

Read that again, because it quietly upends how most senior living operators are spending. 

The trouble is, the 70% doesn’t photograph well. Nobody puts a general ledger on the front of the brochure. And in an industry that lives and breathes its residents, the money and the attention drift, almost gravitationally, towards the AI you can see: the clever fall sensors, the companion tech, the shiny things that win hearts on a tour. 

Meanwhile the plumbing that actually decides whether any of it pays? Left to rattle along on spreadsheets and good intentions. 

This piece is about that plumbing. Not because it’s glamorous, but because in 2026 it has quietly become the thing your investors are staring at. 

First, the uncomfortable number nobody’s putting in the board deck 

Every operator has a slide on AI these days. Adoption rates, pilot counts, the obligatory nod to “innovation.” All very reassuring. 

Here’s the slide you won’t see. 

• The confidence gap. Ask finance leaders, privately, whether they actually believe their AI is working, and the mask slips. Just 36% of CFOs say they’re confident in their ability to drive real enterprise AI impact. Nearly two in three finance chiefs, the very people signing the cheques, quietly aren’t sure the money is doing anything at all. That’s not scepticism from the cheap seats; that’s the person who approved the budget hedging their bets. 

• The retreat nobody names. You’d assume that after two or three years of pilots and enthusiastic all-hands emails, firms would be climbing the maturity curve. They’re not. Today, only 10% describe their AI as “fully scaled and continually evolving” — down from 25% the year before. Read those two numbers together and something unsettling emerges: this isn’t a slow march towards competence, it’s a retreat. Companies are quietly walking their AI back, unwinding the very deployments they were boasting about eighteen months ago, because the returns never turned up. 

• The wrong instinct. When AI underwhelms, the reflex is to assume you bought the wrong tool. Too basic, not clever enough, time to shop around for something shinier. Almost always, that instinct is wrong. As BCG put it rather memorably, “AI in finance fails not because the models lack capability but because the data lacks context”. The engine was never the problem. You didn’t buy a bad car; you poured it onto a road full of potholes and then wondered why the ride was so bumpy. 

• Why the numbers won’t budge. This is why the failure rate refuses to shift no matter how good the technology gets. Each new model is more capable than the last, and each one lands on the same unglamorous, unfixed foundation. The 30% keeps improving. The 70% keeps being ignored. And the results, stubbornly, keep disappointing. So the real question was never “which AI should we buy?” It was always “what on earth are we standing it on?” 

The bias nobody names: senior living funds the AI it can see 

Let’s be honest about something. 

Given the choice between two AI projects, one that helps keep a resident safe and one that tidies up the month-end close, which one wins the room? Every time, it’s the first. And quite right too, in a way. It’s a lovely thing to back. It photographs beautifully. It tells a story your families understand and your board can feel proud of. 

The trouble is, that instinct, warm and well-meaning as it is, quietly bends where the money goes. 

And look where it’s bending. The sector’s energy is pouring almost entirely towards the resident’s door: 

• The market for AI in elderly care was worth about $54 billion in 2025, and it’s on course for $158 billion by 2030. 
• Fall detection, companion tech, predictive care, that’s where the excitement sits. 
• Sage alone raised $65 million in a single round just to grow its fall-prevention AI. 

Now, here’s the bit that ought to give a finance chief pause. 

The research says the money is chasing the wrong door. 

When someone finally sat down and measured this properly, they found that roughly half of all AI budgets go to sales, marketing and the visible, front-of-house stuff, while the quiet back-office work delivers two to three times the return. Read that again. The least glamorous corner of your business, the invoices and reconciliations and reports, is where the actual money hides. And it’s the corner nobody wants to fund. 

Why does this keep happening? It isn’t stupidity. It’s human nature. 

Care tech makes people feel something. A ledger doesn’t. You can’t put a reconciliation engine on the front of a brochure, you can’t show it off on a tour, and it will never once make a family smile. So it waits at the back of the queue, year after year, while the shiny things go first. 

And the cost of that neglect is not hidden, if you care to look. While the clever sensors hum away upstairs, most finance teams downstairs are still “relying on spreadsheets, emails, and manual reconciliation” to get through the month. The future on one floor, the fax machine on the other. 

It isn’t that the sector won’t spend on technology. It spends plenty, tech now eats up close to 10% of capital budgets. The problem is where the spending points. As Ziegler’s own analyst rather neatly put it, the real challenge in 2026 is no longer access to these tools but “discernment”, knowing which ones genuinely earn their keep. 

Spending isn’t the issue. Aiming is. 

So the uncomfortable truth is this. It was never a budget problem. It was a taste problem. We back the AI we can fall in love with, and finance, bless it, has never been much of a looker. 

So what exactly is this “boring stuff”? 

Fair question, because “fix your foundations” is the sort of thing consultants say and nobody acts on. Let’s make it plain. When we talk about the neglected 70%, we really mean three unremarkable things. 

The data. In most operators it lives in separate little kingdoms, payroll here, scheduling there, the care records in one system, the property and the ledger in another. Each one holds a piece of the truth, and none of them talk. Ask two systems what happened last Tuesday and you’ll often get two different answers. 

The process. The way you close the books, chase a payment, or push through a rate change. In too many groups, every community does it slightly differently, in its own homespun way, which means there’s no single, tidy method for a machine to follow. 

The people. The ones who can look an AI in the eye, sense when its answer smells wrong, and push back. This is the hardest gap of the lot. Ask finance chiefs what worries them most right now and building this very skill inside the team tops the list. 

None of it is exciting. None of it will ever trend. But this, and not the model, is the thing that decides whether your AI ever earns a penny. 

Get it right and the clever tools finally have something solid to stand on. Skip it, and you’re simply buying a faster way to be wrong. 

On a related note, if the capital angle in all this caught your eye, there’s a companion read worth your time: The UK Senior Living Supply Gap Is Becoming a Capital Problem. It picks up where this one leaves off, why the money’s there, yet the deals so often aren’t. 

Why senior living is the worst-case environment to skip the 70% 

Here’s the thing most AI advice quietly ignores. It’s written for a tidy company. One head office, one system, one version of the truth. 

Senior living is nothing like that. 

A multi-community operator isn’t a single business. It’s twenty, or forty, or a hundred little businesses stitched loosely together, each with its own quirks, its own habits, and its own way of counting things. And that changes everything about what happens when you add AI to the mix. 

BCG said it about as plainly as anyone could: 

“Automating a fragmented process scales fragmentation rather than eliminating it.” 

Sit with that for a second, because it’s the whole game. If your close is a bit of a muddle in one community, automation doesn’t tidy it up. It simply does the muddle faster, in every community at once. You don’t fix forty messes. You scale them. 

And this isn’t theory. It’s already happening. 

Take Anthem Memory Care. They rolled out AI in good faith and ran straight into trouble, the tools threw up “hallucinations” and served up “conflicting information on the same data sets,” which then “complicated the company’s process used to create pro formas”. Think about what that actually means on the ground. The numbers they use to plan the business, quietly poisoned by a tool meant to sharpen them. 

Now hold that against a very different story. 

Over at Ascent Living, they took the slow road. Since 2021 they’ve been, in their own words, “documenting everything so the AI has enough data to work properly.” Guardrails first. Foundations first. The unglamorous 70% first, and only then the clever stuff on top. One operator reached for the shiny model. The other built the road before letting the car out. Guess which one is actually getting somewhere. 

There’s one more wrinkle here that the generic advice never mentions. 

Your data is bound by rules the tech world simply doesn’t have. You can’t just hoover up resident and staff records and feed them to a hungry model, HIPAA sees to that. So the very foundation senior living most needs to fix is also the one it’s least free to touch carelessly. Harder problem, higher stakes, less room for a shrug. 

Put it all together and the picture is uncomfortable but clear. The sector that most wants AI to work is, by its very shape, the sector where standing it on shaky ground does the most damage. 

The quiet problem hiding underneath: shadow AI 

Now for the part that should genuinely keep a finance chief up at night, because it’s happening right now, under the radar, whether you’ve sanctioned it or not. 

When people aren’t given tools that work, they don’t down tools. They find their own. 

• Roughly half of all employees are already using AI tools their organisation never approved. 
• Around two-thirds have used AI at work despite believing it was against the rules. 
• More than a third have fed customer or financial data straight into public models. 

Read that back in a senior living context and it should make you wince. Somewhere in your finance team, right now, someone frustrated by a clunky spreadsheet may well be quietly pasting sensitive numbers into a free tool, just to get their evening back. 

And 2026 has raised the stakes on this considerably. We’ve moved past chatbots that merely answer questions. The new tools act, they can reach into connected systems and do things on your behalf . An unapproved tool used to be a leak. Now it’s a stranger with a set of keys. 

But here’s the reframe that matters, and it’s easy to miss. 

Shadow AI isn’t really a discipline problem. It’s a symptom. 

People only go rogue when the proper path is too painful to bother with. Every unsanctioned tool in your finance team is really a little flare going up, a signal that the 70% is starving, and that your people have quietly given up waiting for you to feed it. 

Putting a real figure on the “boring stuff” 

Everything so far has been a touch philosophical. Foundations, discipline, and taste. All very well. 

But finance chiefs don’t move on philosophy. They move on numbers. So let’s put a price on the thing nobody wants to fund, because once you see it in pounds and pence, the whole argument stops being a nice idea and starts being a leak in the hull. 

Start with the humblest object in the building. A single invoice. 

Processed by hand, the way most communities still do it, an invoice costs around $9.87 to push through. Automated, that same invoice costs $2.81. Not a dramatic gap on one piece of paper, admittedly. But you don’t process one, do you? And here’s the sting in the tail for an operator with lots of sites: 

“A multi-facility operator with paper-and-email AP sits at the expensive end, because every facility multiplies the formats, the approvers, and the exceptions.” 

Every community you add doesn’t just add invoices. It adds another way of doing invoices. The mess compounds. 

Now widen the lens to the month-end close. 

One study followed a modest three-property operator and found it pouring more than 200 staff hours a month into manual close work, keying numbers, chasing discrepancies, and cobbling reports together in spreadsheets. The bill for that? Roughly $8,000 a month, or $96,000 a year, in wages alone. Ninety-six thousand pounds’ worth of clever, capable people, spent doing something a machine could do in its sleep. 

And that’s just the work you can see. There’s a quieter, nastier cost underneath it. 

Revenue you earned but never actually billed for. It happens all the time in this sector. A care level changes, a new service starts, a rate goes up, and somewhere between the care team, the front desk and the ledger; the information simply falls down a crack. The money was yours. You just never collected it. As one analysis puts it, the cash gets “trapped between resident management, care, occupancy, billing, and accounting systems”, and quietly evaporates. 

Stack these up and a rather sobering shape emerges: 

• 55% of healthcare payments still crawl through manual channels. 
• Cleaning that up can cut processing costs by up to 30% and speed cash collection by 20%. 
• And operators who actually did the unglamorous work have unlocked over $0.5 million in annual savings, some 6,000+ hours, with turnaround times slashed by more than 90%. 

Half a million pounds. From tidying up the boring stuff. That’s not a rounding error, that’s a wing of a new building. 

So the next time an AI project promising to “reimagine the resident journey” is competing for budget against a dull little proposal to automate accounts payable, remember which one quietly hands you back half a million a year. 

The glamorous option wins hearts. The boring one wins the P&L. 

The reason this suddenly can’t wait 

You might reasonably say: fine, but we’ve muddled through for years. Why is this now urgent? 

Because the maths of running these communities has quietly turned against you. 

On paper, business looks grand. Occupancy has climbed to 90%, capping off twenty straight quarters of growth. Rooms are full. Demand is roaring. You’d think everyone would be swimming in margin. 

They’re not, and here’s why. 

The costs have galloped ahead of the good news. Staff pay now swallows an enormous 56% of operating budgets, and a striking 90% of operators say their staffing bill went up this year. Full houses, thinner margins. It’s a strange, cruel sort of squeeze. 

And this is the whole point. 

When you can’t fatten your margin by filling more rooms, because the rooms are already full, the only margin left to find is in how tightly you run the place. In seeing your costs clearly, forecasting the wage bill before it bites and plugging the leaks. 

Which is precisely the work the neglected 70% was built to do, and precisely the work a clever tool on a broken foundation can never manage. 

The boring stuff was always worth doing. In 2026, it’s the difference between a full house that thrives and a full house that quietly bleeds. 

The plot twist: in 2026, the boring stuff is a valuation lever 

Right, this is where it gets genuinely interesting. Because everything we’ve talked about so far, the messy data, the manual close, the leaks, you could file under “operational headache.” Annoying, costly, but ultimately your problem, tucked away behind the scenes. 

Not anymore. 

Something has shifted in the way senior living is owned, and it has quietly dragged your back office out of the shadows and onto the balance sheet. Most operators haven’t fully clocked what it means yet. The ones who do will have a real edge. 

Let me explain. 

For years, the big property owners, the REITs, mostly kept things simple. They owned the bricks, leased them to you on a fixed rent, and left you to run the place. Whatever chaos lurked in your finance function was, frankly, none of their concern. They got their cheque either way. 

That arrangement is unravelling fast. 

Across 2026, nearly every major healthcare REIT has swung back into what’s called a RIDEA or SHOP structure, Welltower, Ventas, Healthpeak, LTC, Omega, the whole cast . The names don’t matter. What matters is the shift underneath them. 

Under the old way, the owner collected rent. Under this new way, the owner shares directly in how well, or how badly, each community actually performs . Your operating profit is now their operating profit. Your leaks are now their leaks. 

Do you see what that does? 

It takes the quality of your finances, once a private matter, and turns it into the very thing your capital partner is buying. 

Which is why the same analysis landed on a line that should stop every operator in their tracks: 

“Operator selection is now the #1 risk variable in senior housing investment underwriting.” 

Read that slowly. Not the location, not the building, not the demographics. You. How you run the place, how clean your numbers are, and how much an investor can trust the figures you hand them. 

And make no mistake, the money is watching closely, because there’s an awful lot of it about. 86% of investors say they want more senior housing on their books in 2026 , chasing an over-80 population set to swell by nearly 37% in a decade , with some $24 billion in deals already done . The appetite is enormous. The question every one of those investors is quietly asking is the same: whose numbers can I actually believe? 

Here’s the part that turns this from a worry into an opportunity. 

Clean operations aren’t just safer. They’re worth more, in hard cash. Communities running on solid, well-fed operational data have posted something like $275 more in profit per resident, every single month . Multiply that across a portfolio and you’re not talking about a tidier spreadsheet. You’re talking about a materially more valuable business. 

So let’s retire the word “overhead” for a moment. 

Your back office isn’t a cost your investor grudgingly tolerates. Under this new model, it is the return they’re buying. The cleaner it is, the more your business is worth, the cheaper your capital, the easier every future deal becomes. 

Which brings us to the line I’d want on the wall of every senior living boardroom this year: 

The businesses that win the next capital cycle won’t have the flashiest AI. They’ll have the cleanest foundations for it. 

The fall sensors will still win the tour. But it’s the reconciliation engine, dull, invisible, endlessly neglected, that increasingly decides what your company is actually worth. 

Funny old world, isn’t it. 

A CFO’s re-sequencing playbook 

So what does a finance chief actually do with all this on Monday morning? 

Not rip everything out. Not freeze AI spending and wait for a perfect world that never arrives. The fix here isn’t dramatic, it’s simply a matter of changing the order in which you do things. Getting the sequence right. Here’s where I’d start. 

1. Fund the invisible before the visible. The next time a dazzling resident-facing project is squaring up against a dull little back-office one for the same budget, pause. Back the dull one first. Not forever, just first. And you don’t need to boil the ocean either, pick one painful, measurable job, the close, or accounts payable, or the wage forecast, and fix that. As BCG rather wisely notes, “momentum matters more than perfect data” . One solid win funds the next. 

2. Tidy the process before you automate it. This is the one everyone skips, and it’s the one that bites hardest. If every community closes its books its own homespun way, automating that simply means doing forty different muddles at speed. So before a machine touches anything, agree on one way. One close. One approval flow. One method. Standardise first, automate second. Never the other way round. 

3. Make everyone speak the same financial language. Remember that “resident day” that meant two different things in two different systems? Multiply it across a portfolio and you’ve got an AI confidently producing nonsense, because it’s reasoning on words that don’t line up. Settle your definitions. One chart of accounts, one shared vocabulary, so that “margin” means the very same thing in every building. Dull work. Utterly essential. 

4. Give your people proper tools, and the nous to use them. Two jobs here, really. 

• First, kill off shadow AI, not by banging the rule book, but by making the sanctioned path the easier one. People go rogue when the official route is a pain; take the pain away and the problem mostly solves itself . 
• Second, build the human judgement to sit alongside the machine. This is the skill finance chiefs themselves say they’re most worried about right now . You want people who can look at an AI’s answer and smell when something’s off, not simply wave it through. 

5. Decide, deliberately, what stays human. Not everything should be handed to a machine, and pretending otherwise is how you end up in trouble. The audit trail, the controls, the covenant maths, the judgement calls a regulator might one day ask about, keep those in human hands. Let AI do the heavy lifting. Let people own the things that carry consequences. 

Do these in order and something rather nice happens. The foundation firms up, the clever tools finally have somewhere solid to stand, and the returns that never used to show up quietly start to. 

A quick five-question audit for your next board meeting 

Before you approve another penny of AI spend, try these on for size. If any one of them makes you shift in your seat, you’ve found where the 70% is hiding. 

• Does “margin” mean exactly the same thing in every community’s books, or does each site tell its own story? 
• Could an AI trace a month-end surprise back through your data on its own today, or would it need a human to translate first? 
• Was your last approved AI project something you’d proudly show an investor, or just something that looked good at the offsite? 
• Do you actually know how many unapproved AI tools your finance team is quietly leaning on right now? 
• If a capital partner put your back office under the microscope tomorrow, what would they knock off your valuation for? 

Sit with the ones that stung. That’s your to-do list. 

What’s the Bottom Line? 

Let’s bring this home. 

The clever model, the slick platform, the thing that dazzles in the demo, that’s the 30%. It’s real, it’s useful, and it’s the easiest 30% in the world to fall for. But it was never where the money lived. 

The money lives in the other 70%. The data that agrees with itself. The process that runs the same in every building. The people who know when to trust the machine and when to raise an eyebrow. Unglamorous, every bit of it. And decisive. 

Here’s the shift worth carrying into your next leadership meeting. For years the question was“which AI should we buy?” It’s the wrong question, and it always was. The better one, the one that actually separates the winners from the disappointed, is quieter: 

“Is the ground ready for it? And would an investor pay for what’s underneath?” 

Because that’s the real change in 2026. The boring stuff stopped being back-office housekeeping and quietly became the thing that decides what your business is worth. Get it right and everything downstream gets easier, the AI works, the margins hold, the capital comes cheaper and stays longer. Skip it, and no model on earth will save you, you’ll simply have bought a faster way to be wrong. 

The fall sensors will still win the tour. Let them. But do spare a thought for the reconciliation engine humming away in the back, because in this market, that’s the thing your investors are really buying. 

Sort out the foundations first. The clever stuff has waited this long, it can wait a moment more. 

If You’d Rather Not Fix the 70% Alone 

And if that groundwork feels daunting, that’s rather the point of having a specialist in your corner. Sorting out the 70%, the clean data, the standardised close, the finance function an investor can trust on sight, is precisely the sort of unglamorous, high-stakes work QX Global Group does for senior living operators day in, day out. It’s worth knowing what that partnership tends to look like in practice: 

• Consistent processes across every community, so one messy close doesn’t get quietly multiplied across the whole portfolio. 
• Investor-grade reporting and cleaner cost, margin and asset-level visibility, the kind of numbers that hold up the moment due diligence begins. 
• Tighter AP, AR, reconciliations and month-end, with automation and standardisation doing the heavy lifting rather than more headcount. 
• Finance support that scales as the portfolio grows, and, for operators wrestling with the talent gap, access to a deep bench of sector-specific accountants, with cost savings that tend to land somewhere in the 40 to 60% range. 

None of it will ever make the brochure. But it’s the groundwork that lets all the clever stuff finally pay its way. 

If you’re weighing up where your own 70% quietly stands, it might be worth a conversation. Book a consultation with QX, no pitch, no obligation, just an honest look at where your foundations are solid and where they could do with some shoring up before your next AI project, or your next investor, comes knocking.