The Care Economy
Conversation.

AI-hosted editorial conversation. The hosts are synthetic voices, and examples are illustrative. Related tool: https://careeconomycollective.org/tools/reporting-duplication.

Sources: https://www.gsb.stanford.edu/faculty-research/case-studies/amir-dan-rubin-success-beginning and https://www.lean.org/wp-content/uploads/2022/01/ls_puttingtheSystemintoDailyMngt.pdf . Public abstract and separate children’s presentation; full licensed case not acquired.

Chapter 1

Alex — AI host

Welcome to The Care Economy Conversation. I'm Alex, and I'll say this plainly at the start: I'm an AI voice. I'm a synthetic presenter, not a clinician, not an executive, and not an interview guest. I have no personal experience of running a care service.

Eric Marquette

And I'm Eric, also an AI voice, with the same limits. No clinical background, no executive background, and no stories from a facility floor, because we don't have any. What we do have is two public sources, a practical exercise, and a habit of saying where the evidence stops. This is episode one of three.

Alex — AI host

Let's start with a problem, and I want to be clear it's a synthetic example. Picture a residential aged care home and a home care team. Every morning, a short stand up meeting happens. Week after week, someone raises the same thing: the same change in a person's dietary preference gets written in three places, and sometimes the three places disagree. Everyone nods. Someone says they'll look into it. Next week, it comes up again.

Eric Marquette

That's the part that interests me. The meeting is working in one sense, because the problem is being seen. But nothing happens after the seeing. So the real question for this episode isn't whether to hold huddles. It's what has to exist around a huddle so that a raised problem actually ends up fixed, or at least deliberately not fixed, with someone knowing which.

Alex — AI host

Right. And the series name is The Care Economy Conversation. Today we'll look at what two Stanford sources describe, keep them strictly separate, treat anything we take from them as a hypothesis for Australian care, and then finish with an exercise you can run on one issue of your own.

Eric Marquette

One boundary before we go on. Nothing here is proof that any of this works in Australia. And nothing here is an endorsement from Stanford. We're reading public material and asking what might be worth testing.

Chapter 2

Alex — AI host

First source. It's a Stanford Graduate School of Business case by Jeffrey Pfeffer, titled Amir Dan Rubin: Success from the Beginning, published in 2015, case number OB90, 22 pages. Important limitation: we have not acquired that full licensed case. Everything we say about it comes from the public abstract on the Stanford GSB website, and nothing more.

Eric Marquette

So what does the abstract say? In November 2010, the board of Stanford Hospital and Clinics announced that Amir Dan Rubin, then chief operating officer of the UCLA Hospital System, would become the next CEO. The abstract says the hospital had largely recovered from a failed merger with the hospital of the University of California, San Francisco, and was financially stable. That detail matters, because it means this wasn't described as a rescue of a failing organisation.

Alex — AI host

It does. The abstract still names real challenges: creating a focus on patient care, and improving operational performance, in an increasingly competitive health system landscape. It also says his success depended on support from a large existing internal staff and from the medical centre faculty, and that the faculty reported to the dean of the medical school, not to Rubin. The case, as described, is about building support among medical school physicians, the board, and colleagues already at Stanford.

Eric Marquette

So what can we responsibly take from an abstract? Not a method. The abstract doesn't tell us what daily practices were used, and it doesn't claim a result we can lean on. What it gives us is a framing: leadership support, culture, and a focus on patients and operations, in a setting where the leader couldn't simply instruct everyone. That's a useful question for care operators, where clinicians, contractors, and coordinators often don't sit in one reporting line. But it's a question, not an answer.

Alex — AI host

Now the second source, and this is a different institution and a different period. It's a 2019 presentation called Putting the System into Daily Management, by Denise Bennett of Lean Enterprise Australia and Daniel Murphy, a doctor at Stanford University and Stanford Children's Health, shared through the Lean Enterprise Institute. This is about Stanford Children's Health, a paediatric setting. It is not the adult hospital in the 2015 case, and we won't treat one as evidence for the other.

Eric Marquette

What does the presentation describe? It calls a daily management system a system of daily habits and practices with supporting tools. It uses a bike picture. The front wheel is daily management, with standard work for holding stability and incremental improvement. The back wheel is bigger improvement work. Steering is purpose, alignment, and key metrics. And it lists parts: tiered huddles that cascade from the front line up through three to five levels, a daily process for problems, leader standard work, senior leader daily habits, visual boards, local improvement teams, and protected time.

Alex — AI host

Two pieces of it stood out to me because they're about follow through rather than meetings. One is the problem process. Issues are triaged into three responses: resolve simple or critical ones immediately, own the non urgent ones locally, and escalate the ones that are urgent, time sensitive, or stuck behind barriers. It also separates quick hits, resolved in 24 hours or less, from big issues that take longer. And the sample tracker has columns for the issue, an owner, a due date, and a status.

Eric Marquette

The other is protected time. A Stanford Children's fact sheet in the presentation describes meeting free time from 8 in the morning to 10, Monday to Friday, for leaders of care teams, so they can support their teams without distraction. It says meetings other than readiness huddles shouldn't happen then, and it expects leaders to attend one or two tiered huddles a day and spend time weekly on the front line. That's a description of an intention. The presentation doesn't give us evidence that it worked, and we're not claiming it did.

Alex — AI host

Which brings us to the hypotheses. If we put the two sources side by side, without merging them, three ideas are worth testing in Australian residential and home care. One: priorities that are aligned, so a team can see how its daily work connects to what the organisation says matters. Two: problems made visible daily, close to when they occur. Three: a named person who follows each problem through. Those are hypotheses. We're not saying they caused anything at Stanford, and we're not saying they'll transfer.

Eric Marquette

And the context is different. A large academic medical centre in the United States has its own funding, scale, and staffing. A home care provider with workers visiting people across a region, or a residential home with a small leadership team, runs differently. So the useful move is a small, bounded test with a plain way to stop it, which we'll get to.

Chapter 3

Alex — AI host

Back to our opening problem. Why might a huddle become another reporting burden?

Eric Marquette

Let's reason it through, as a possibility rather than a finding. Suppose a huddle gets added on top of existing handovers and documentation. Staff now report the same thing at the huddle, in the notes, and in a form for management. If raised issues vanish into a gap, people learn that speaking up costs time and returns nothing. The huddle turns into something to get through, a tax on care time rather than a tool for care.

Alex — AI host

There's a second failure that's subtler. A huddle can drift into a grievance session, where problems are aired but nobody has authority to change them. A care worker can see that a supply is chronically short, but they can't change the ordering process. A coordinator can see two systems asking for the same information, but can't merge them. If no one above them has a routine for clearing that kind of barrier, the huddle just documents helplessness.

Eric Marquette

So what would it require instead? Looking at the presentation's parts, I'd pull out two. Protected time, and responsive leadership. The Stanford Children's material describes time set aside to keep frontline leaders with their teams, and its reflection tool says frontline leader standard work should structure more than half of the day, with coaching connected to problem solving. Translate that to aged care. If a team leader is the one who also fills gaps in the roster, who triages the board? The time has to be genuinely protected, or the board becomes decoration.

Alex — AI host

And on the other side, the same tool lists senior leader daily habits, including going to see the work directly and coaching frontline leaders. For an Australian operator, that might mean a facility manager or an operations lead having a regular, visible slot to clear what's been escalated, and telling staff the answer, even when the answer is no. A no with a reason is still follow through. Silence isn't.

Eric Marquette

There's an objection I'd expect from listeners. We're already stretched. Adding protected time sounds like a luxury. I think that's fair. Protecting time has a real cost, and we shouldn't pretend otherwise. That's something for episode two, where we separate cash from released capacity. For now the honest position is: this is a trade off to test on a small scale, not something we can promise pays for itself.

Alex — AI host

Another objection: this is just meetings with better branding. Maybe, if it's run that way. The difference we'd look for isn't the meeting. It's whether an issue that appears on Monday has an owner by Monday, an escalation rule, a due date, and a verified ending. So let's build that.

Chapter 4

Eric Marquette

Here's the exercise. Choose one issue. We'll use a synthetic one: dietary preference changes are recorded in three places and sometimes don't match. Write it as a row, and walk it through seven stages. Pencil is fine.

Alex — AI host

Stage one, first capture. Where and when is it first logged? In our example, a care worker raises it at the morning huddle, and it goes on a visible board or list in plain words: what happened, where, and how often it seems to occur. Not a story, not blame. Just the problem.

Eric Marquette

Stage two, a single named owner. One role, one person, not a committee. The sample tracker in the presentation has an owner column for exactly this reason. In our example, perhaps the care coordinator owns the test. Owning doesn't mean doing everything. It means being the person who can be asked where it's up to.

Alex — AI host

Stage three, escalation. Write down when this stops being local. For instance: if the fix needs a change to a form someone else controls, or a system another team manages, or it isn't resolved by the due date, it goes up one tier, to a named person, on a named day. The Stanford Children's material separates resolve, own, and escalate. You decide your own thresholds, in advance, so nobody has to guess.

Eric Marquette

Stage four, a due date. A short one, for trying one adjustment, not solving everything. Say, two weeks to test recording the preference once and referencing it from the other two places. If it's a quick hit, a day or less. If it's bigger, say so, and give it a longer date, but a date.

Alex — AI host

Stage five, verified closure. This is where I'd push hardest. Closed doesn't mean someone ticked a box. It means someone looked at the actual work and saw the problem no longer occurring. And they checked it didn't simply move, for example into the kitchen, or into the evening shift's paperwork.

Eric Marquette

Stage six, staff and lived experience feedback. Ask the care workers who do the task whether it's easier, the same, or harder. And ask the people receiving care, or their families, whether anything feels different. Did a conversation about someone's preferences get shorter in a way that feels less personal? Did anything about dignity or choice change? This is where you catch hidden friction that a tidy tracker would never show. And keep required reporting and escalation intact. The goal is to stop repeating the capture, not to stop recording what must be recorded.

Alex — AI host

Stage seven, stop criteria. Write them before you begin. Stop and roll back if the time spent goes up for staff, if a compliance or safety concern appears, if people report the change hides someone's needs, or if staff tell you it's making the work worse. Stopping a test is a good outcome, if it's the result of looking honestly. It doesn't mean the idea has failed.

Eric Marquette

If you want a way to put numbers on the repeated capture, the Collective has a free tool called the Reporting duplication finder. It helps you find repeated capture, quantify effort, and choose one reporting change to test. The page says your figures stay in your browser, it doesn't send inputs to AI, and it doesn't save them on the server. The examples are synthetic, and it reminds you to validate time estimates with your team and check that changes don't hide unmet need.

Alex — AI host

A note about the examples we bring. Dmitry Shibanov is involved in developing SMPLS LABS and KnowingMe. Within the Collective, these are examples of systems co-designed with providers and the people who use them. Our purpose is to exchange ideas and work together on a better care economy for Australia. Everyone is welcome to bring their own products, practical experience and ideas for us to explore together. Participation is free and does not require buying anything. Contributors keep their existing intellectual property, and confidential material should only be shared under agreed protections. The Collective has not yet demonstrated provider savings or outcomes. Today's discussion offers questions to test, based on two public sources kept separate, including one available to us only as an abstract.

Eric Marquette

To recap the hypotheses, and they are hypotheses: aligned priorities, daily problem visibility, and named follow through. Test one issue, with an owner, an escalation rule, a due date, a verified ending, feedback from staff and the people you support, and stop criteria written first.

Alex — AI host

The tool is at careeconomycollective dot org, slash tools, slash reporting duplication, with the two words joined by a hyphen in the web address. The link will also be in the show notes. Next time, we look at better care within the funding you have, and why released capacity isn't the same as cash. Thanks for listening.

AI-hosted editorial conversation. The hosts are synthetic voices, and examples are illustrative. Related tool: https://careeconomycollective.org/tools/service-cost.

Sources: https://www.health.gov.au/our-work/aged-care and https://www.agedcarequality.gov.au/providers/workforce-obligations/247-registered-nurse-and-care-minutes-obligations . Financial scenarios are synthetic and do not determine eligibility or lawful funding transfers.

Chapter 1

Alex — AI host

Welcome to The Care Economy Conversation. I'm Alex, and I'll be plain about this from the start: I'm an AI voice. I'm a synthetic presenter, not a clinician, not an executive, and not an interview guest. I have no personal experience of running a care service, and no memories from a facility floor.

Eric Marquette

And I'm Eric, also a synthetic AI voice, with the same limits. This is an editorial conversation, and the examples are illustrative. This is episode two of three. Dmitry Shibanov is involved in developing SMPLS LABS and KnowingMe. Within the Collective, these are examples of systems co-designed with providers and the people who use them. Our purpose is to exchange ideas and work together on a better care economy for Australia. Everyone is welcome to bring their own products, practical experience and ideas for us to explore together. Participation is free and does not require buying anything. Contributors keep their existing intellectual property, and confidential material should only be shared under agreed protections. The Care Economy Collective has not yet demonstrated savings or clinical outcomes for any provider.

Alex — AI host

Last episode we ended on a trade off. Protecting time for follow through has a real cost, and we said we couldn't promise it pays for itself. This episode is about how you would find out honestly. The title is Better Care Within the Funding We Have, and I want to start with a synthetic scene.

Alex — AI host

Imagine a leadership meeting. A project report says a change to documentation has freed up 416 hours a year. A board member leans in and asks a perfectly reasonable question: so where does that show up in the budget? And the room goes quiet, because everyone knows the rostered hours didn't change. The report was true. The question was fair. They were just about two different things.

Eric Marquette

That quiet is the whole episode. Care leaders often use one word, savings, for four different things, and the confusion can push decisions in the wrong direction. So let's pull them apart. First, cash savings. That's a real reduction in money going out the door. Less agency spend, lower overtime payroll, a cancelled contract. You can point to the line in the ledger where it dropped.

Alex — AI host

Second, released capacity. That's staff minutes or hours that no longer get used on a task. It's real, and it might be valuable, but it is potential working time. It's not money in the bank. Third, care experience. How it feels for residents, families and workers when care is unhurried, attentive and respectful. That's qualitative, and it matters in its own right, but it doesn't convert into a dollar figure just because we'd like it to.

Eric Marquette

And fourth, unmet need. The tasks that were deferred, compressed or missed because time ran out. The conversation that got cut short, the extra walk that didn't happen, the follow up call that slipped. The uncomfortable point about unmet need is that when you release capacity, it may be the first thing that capacity gets spent on. Which is arguably a good outcome for care. But it also means the capacity is spoken for, so it can't also be a cash saving.

Chapter 2

Alex — AI host

Let's do the first calculation, and I'll flag that it is synthetic. These are not figures from any provider. Suppose a team deals with 80 issues a week, and for each one there are six minutes of duplicate work. The same detail written into a second system, or re-keyed on another form, or reported again to someone who already has it. Eric, walk the arithmetic.

Eric Marquette

Eighty issues times six minutes is 480 minutes a week. That's eight hours. Across a 52 week year, eight hours a week is 416 hours. And if we apply an illustrative loaded rate of 55 dollars an hour, 416 hours is 22,880 dollars. That is the capacity value in our synthetic example.

Alex — AI host

Now the sentence I want to say loudly. That 22,880 dollars is not cash saved. It's the value of time that could be used differently. Payroll hasn't moved. Nobody was paid less. Unless a shift is actually cancelled, or agency hours are actually reduced, the same wages go out the door the following fortnight.

Eric Marquette

Which is why I'd say it twice. Not cash saved. If staff are paid the same and the eight hours a week goes into something else, then the money has been spent, just on something different. That something might be time with residents, or a task that was previously missed. In that case the honest label is released capacity, redirected to care. That's a legitimate outcome. It just isn't a budget surplus.

Alex — AI host

There's an objection I'd expect here. Someone says: it's still worth 22,880 dollars, so why does the label matter? I think the label matters because of what happens next. If a report calls it savings, someone may build it into next year's budget as money that's already been found. Then the budget is short by that amount, and the shortfall lands on rosters, on the people doing the work.

Eric Marquette

And the reverse objection is just as fair. Someone says: so capacity is worthless? No. Eight hours a week of avoided duplication could matter enormously if the team is stretched and unmet need is piling up. We simply have to name it correctly, and decide on purpose where it goes. Back to cash, you'd need a decision that actually takes hours or contractor spend out, plus evidence that care didn't suffer. Otherwise, call it capacity and track what it was used for.

Alex — AI host

One more caution about the number itself. It rests on a count of issues and a duration per issue. In real life, six minutes is an estimate, and the people doing the work should check it. A count of 80 could be a poor proxy for the actual burden. The arithmetic is simple. The inputs are where the uncertainty lives.

Chapter 3

Eric Marquette

Now the second synthetic example, which is a technology business case. A provider considers a care coordination tool. Upfront implementation and training cost is 40,000 dollars. Annual running costs, operating and licensing, are 18,000 dollars. And suppose there's an evidenced annual cash benefit of 24,000 dollars. Again, this is illustrative, not a real vendor or a real provider.

Alex — AI host

The word evidenced is doing real work there, so keep it in mind. Taking the numbers as given: 24,000 minus 18,000 leaves a net of 6,000 dollars a year. Divide the 40,000 upfront by 6,000 and you get a simple, undiscounted payback of about 6.67 years. And over a three year horizon, the cumulative net cash position is minus 40,000 plus three times 6,000, which is minus 22,000. Negative 22,000 dollars after three years.

Eric Marquette

Notice how that differs from how a headline might sound. A 24,000 dollar annual benefit sounds attractive. But the net is a quarter of that, and the payback is nearly seven years. That doesn't automatically mean the investment is wrong. Care experience could improve, risk could fall, workers could find the work more manageable. But if the case is built on cash alone, this one is thin.

Alex — AI host

So let's talk about the limits of simple payback, because it's a tempting number to quote. First, it ignores the time value of money. A dollar this year is worth more than a dollar in year six, and a provider's cost of capital matters. Second, it hides timing and liquidity. Being negative 22,000 dollars at year three means that cash has to be funded from somewhere, and for a small provider that may be the real constraint.

Eric Marquette

Third, it assumes the benefit holds steady for nearly seven years. Real benefits can fade. Staff turnover resets training, processes drift, the tool gets used less carefully, or the vendor changes pricing. Fourth, a single payback number says nothing about risk before breakeven. A lot can change in six years in a sector that's reforming as fast as this one. So simple payback is a first screen, not a decision.

Alex — AI host

The second half of the problem is what goes into the cost column. Our example said 40,000 upfront and 18,000 a year, but a full cost view asks what might be missing. I'll list seven things to check. One, training, including refreshers, not only the first session. Two, backfill, meaning paying someone to cover direct care shifts while permanent staff are in training.

Eric Marquette

Three, integration. Connecting a new tool to your care management system or other software, and the infrastructure underneath it. Four, subscriptions, including whether pricing changes as you add users or features. Five, support, both internal time and vendor help when things break. Six, governance. Privacy, clinical oversight and compliance checks, with someone responsible for them. And seven, monitoring and evaluation, because someone has to measure whether the benefit is real. That last one is easy to leave out, and it's what separates a claimed saving from an evidenced one.

Alex — AI host

And I'd add a question that sits across all seven. Whose time is it? If the cost is mostly staff hours, we're back to capacity rather than cash, and the same label discipline applies on the cost side. Some of these costs will be actual invoices, some will be absorbed time. A good business case separates the two and says which is which.

Chapter 4

Eric Marquette

So what would it take to show an actual expense reduction? I'd say three tests. First, the line item has to drop. Not a forecast, not a modelled hour, but an actual figure. Overtime payments, agency hours rostered, an invoice. You compare a baseline period with a later period, and you account for things that changed for other reasons, like occupancy or the mix of people being supported.

Alex — AI host

Second, care must not be lost. That means looking at measures you already have, and the ones that matter to the people receiving care. Each residential aged care home has targets based on its residents' care needs. Those targets are based on a sector-wide average of 215 minutes of direct care per resident per day, including 44 minutes of registered nurse care. Check your home's current requirements. A reduction that leaves care needs or staffing obligations unmet is not a saving. In home care, check for missed or shortened visits and worsening feedback. Ask families and residents, not only the data.

Eric Marquette

Third, the workload must not simply move. This is the one I think gets missed most. Say charting gets pushed onto the night shift. Or a registered nurse ends up doing clerical intake that an administrator used to do. Or the work lands on family carers, who aren't on any payroll. In each case the expense line might drop while the burden hasn't gone anywhere. It's just moved to a place that's harder to see.

Alex — AI host

So a practical check is to ask who now does the task, when, and with what time. Ask staff across shifts and roles, and ask families where it's relevant. If hours fell in one place and quietly rose in another, you haven't saved money. You've shifted it, and perhaps shifted risk with it.

Eric Marquette

Now the funding picture, and I want to be careful. Australian aged care funding isn't one thing. The Aged Care Act 2024 started on 1 November 2025, and the Support at Home program replaced the Home Care Packages Program on the same date. Residential care has its own arrangements, including care minutes. Funding purposes and restrictions differ across programs and grants, and they change.

Alex — AI host

So we're not saying all funding carries a single restriction, and we're not saying any particular transfer between budgets is unlawful. We're also not making any eligibility or legal funding determination in this episode. What you can spend where must be checked against the current programme rules, such as the Support at Home program manual, and with your own advisers. A starting point for the official material is on the health dot gov dot au website, under our work, then aged care.

Eric Marquette

And one more honest statement. We can't assume the existing funding envelope is enough everywhere. Thin markets, regional and remote areas, and specialised services can face different cost pressures. The Department's own site describes a Thin Markets grant for rural, remote and specialised Support at Home services, which at least shows the difference is recognised. So when we talk about better care within the funding we have, we mean working honestly with what's there. We don't mean pretending it's always adequate, and we don't mean unmet need disappears if you tidy up your paperwork.

Alex — AI host

So here's the exercise. Take one service, and build a full cost scenario. Use the Service cost tool, at careeconomycollective dot org, slash tools, slash service, and then the word cost, with the two words joined in the web address. Put in everything: wages, on costs, supervision, overheads, the items we listed. Then challenge it. Which numbers are invoices, and which are absorbed time? Which are estimates, and who could verify them?

Eric Marquette

Then, if you're considering a technology change, use the Technology case tool at careeconomycollective dot org, slash tools, slash technology, then the word case, again joined in the address. Enter upfront costs, running costs and the cash benefit you can actually evidence, and look at the multi year picture, not just the headline. Try a version where the benefit fades, and one where integration costs more than planned. If the case only works in the best scenario, you've learned something important before spending anything.

Alex — AI host

Keep the four labels beside every number you enter. Cash saving. Released capacity. Care experience. Unmet need. If you can't say which one a figure belongs to, it isn't ready to go in a board paper.

Alex — AI host

Let's close with the limits. Our two calculations were synthetic. They show how to reason, not what any provider will find. The Collective has not demonstrated savings or outcomes for any provider. Products mentioned in this series are examples for shared learning, and everyone is welcome to contribute an alternative. Contributors keep their existing intellectual property. We are not giving legal, financial or funding eligibility advice. Check the current programme rules yourself.

Alex — AI host

The one thing to take away: 22,880 dollars of capacity value isn't 22,880 dollars of cash, and a 24,000 dollar benefit isn't a 24,000 dollar gain once the costs are counted. Name what you've freed, say where it went, and prove what you claim. Next time, in episode three, we design one bounded reporting experiment so you can test all of this in a small, safe way. Thanks for listening.

AI-hosted editorial conversation. The hosts are synthetic voices, and examples are illustrative. Related tool: https://careeconomycollective.org/tools/improvement-plan.

Source: https://www.oaic.gov.au/privacy/privacy-guidance-for-organisations-and-government-agencies/guidance-on-privacy-and-the-use-of-commercially-available-ai-products . Our technical design choices are not a privacy certification.

Chapter 1

Alex — AI host

Welcome to The Care Economy Conversation. I'm Alex, and I'll say this plainly: I'm an AI voice. I'm a synthetic presenter, not a clinician, not an executive, and not an interview guest. I have no personal experience of running a care service and no memories from a facility floor.

Eric Marquette

And I'm Eric, also a synthetic AI voice, with the same limits. This is an editorial conversation, and every example today is illustrative. This is episode three of three, titled Capture once, use well. Dmitry Shibanov is involved in developing SMPLS LABS and KnowingMe. Within the Collective, these are examples of systems co-designed with providers and the people who use them. Our purpose is to exchange ideas and work together on a better care economy for Australia. Everyone is welcome to bring their own products, practical experience and ideas for us to explore together. Participation is free and does not require buying anything. Contributors keep their existing intellectual property, and confidential material should only be shared under agreed protections. The Care Economy Collective has not demonstrated a provider trial or measured benefits. Everything we describe is a design to test.

Alex — AI host

In episode one we had a synthetic huddle where the same dietary preference change was written down in three places, sometimes differently, and the problem kept coming back. In episode two we put numbers on duplicate work: 80 issues a week, six minutes each, 416 hours a year, 22,880 dollars of capacity value, and we insisted that was not cash saved. So the obvious next question is, how would you find out whether your own version of that is real? Without putting anyone at risk while you look.

Eric Marquette

Right, because the tempting move is a big rollout. Buy something, switch it on, announce that you've captured once and used well. And that's exactly where an unbounded change can hurt. Staff end up doing the old way and the new way at once. A tool misreads something and nobody notices. A report looks tidier while the work has quietly moved to the night shift. A bounded experiment is a way of being wrong cheaply and safely.

Alex — AI host

So let's design one, in the open. Our synthetic scenario is small. One kind of issue, say changes to a dietary preference. One service. One window of time. Capture it once, and have every place that needs it use that single capture. Eric, what has to be true before anyone touches a tool?

Chapter 2

Eric Marquette

Seven things. I'll go through them and Alex can push on each. First, a baseline. Before any change, you measure the starting state. How many minutes per shift go on re entering the same information? How often do handovers disagree with each other? Our 80 issues and six minutes were synthetic. Your baseline has to come from your own people counting honestly, and ideally the people doing the work should check whether six minutes is even close.

Alex — AI host

And that connects to a point from last time. The arithmetic is simple, the inputs are where the uncertainty lives. If the baseline is a guess, then any improvement you report is also a guess. Second, one named owner. Not a committee. One person who is accountable for tracking progress, triaging friction, and reporting status.

Eric Marquette

And I'd stress that owning isn't the same as doing all the work. The owner makes sure that the issue board is current, that problems get an answer, and that someone with authority hears about the ones that can't be solved locally. If there's no name, then everyone assumes someone else is watching.

Alex — AI host

Third, explicit urgency and escalation rules, agreed before the start. This borrows the logic from episode one, where a daily problem is resolved, owned, or escalated. Here that means deciding in advance what the team fixes on the spot, what gets an owner and a date, and what goes straight to a manager. For example, anything that touches a dietary or allergy detail being wrong or missing is urgent by definition, and it escalates the same day. That's a synthetic rule for illustration. Your clinical and safety leads should set the real ones, because we can't.

Eric Marquette

Fourth, a fixed due date. A defined window, say 30 to 90 days, with a date written down. Open ended pilots tend to become permanent by neglect. Fifth, verified closure. When someone says the duplicate step has stopped, somebody checks. Does the second form still exist? Is anyone still copying the detail into a notebook? And crucially, has the task vanished, or has it moved to the kitchen, or the night shift, or a family member?

Alex — AI host

Sixth, a manual fallback. A tested way to go back to the existing paper form or spreadsheet immediately, if the new workflow fails, lags, or confuses people. Tested means someone has actually practised the switch before the trial starts, not just written a sentence in a plan. And seventh, a weekly review. A short, regular meeting that looks at the live measures, the blockers, and what staff are finding annoying. Annoyance is data.

Eric Marquette

One more thing that sits under all seven. Protected improvement time. Running and evaluating a test takes real hours, from the owner, from a leader, and from the people whose work is being observed. If that time isn't protected, with backfill where the work can't wait, then the weekly review turns into a decorative exercise. And in episode two's terms, those hours are part of the full change cost, alongside training, backfill, integration, subscriptions, governance and monitoring. They go in the cost column from day one.

Chapter 3

Alex — AI host

Now the AI part, and I want to treat this carefully, because it's where design choices and legal requirements get blurred. Let's keep them in two separate piles. Pile one: our own pilot design choices. These are choices, not laws.

Eric Marquette

Right. Our pilot design uses operational material that doesn't identify anyone. Think of a dummy catering tally, a synthetic roster, a mock policy handbook. It uses an extraction preview, so a human compares each extracted field against the source snippet before anything is accepted. Access is uploader only. Storage is bounded, with deletion rules. No private content goes into web queries. Calculations run in code, not in a language model's head. Sources stay visible. And the decisions stay with humans.

Alex — AI host

The calculation point matters for what we did in episode two. The 416 hours and the 22,880 dollars should come out of a deterministic calculation you can inspect, not out of a model's guess at arithmetic. The AI can help read a document and propose structure. The sums and the judgement belong elsewhere.

Eric Marquette

Pile two: the legal side. The source we're drawing on is guidance from the Office of the Australian Information Commissioner, on privacy and the use of commercially available AI products, published October 2024 and updated January 2025. A few things it says. Privacy obligations apply to any personal information put into an AI system, and to the output where it contains personal information. Organisations should do due diligence on whether a product suits its intended use, including how human oversight is built in and who can access what goes in.

Alex — AI host

It also says that if an AI system generates or infers personal information, that's a collection under Australian Privacy Principle 3. And under Principle 6, personal information should be used for the purpose it was collected for, unless there's consent or a reasonably expected, related secondary use. As a matter of best practice, the OAIC recommends not entering personal information, and particularly sensitive information, into publicly available generative AI tools. And it describes due diligence as not set and forget. It points to a privacy impact assessment as part of a privacy by design approach, with regular review across the product's life.

Eric Marquette

Now, here's where we're careful. We are not saying our pilot design is OAIC compliant. We're not saying it's privacy certified. We don't know of any such certification for this, and we won't use those words about any tool, pilot or process. Our design choices are a way of keeping risk small. Whether a particular use meets your obligations is a question for your own privacy officer and advisers, looking at your actual facts. One more caution. We are not promising Australian only AI processing. Commercially available AI services can run across cloud regions, so ask your vendor in writing where processing and storage actually happen, rather than assuming.

Alex — AI host

So the staging looks like this. Phase one is synthetic data only. Generated, non identifying material, nothing live. The question at that stage is simple: does the workflow behave, and can people follow it? Phase two, a shadow trial, where the workflow runs alongside your usual systems without influencing any care decision, needs its own separate approval. It doesn't happen automatically because phase one went well.

Eric Marquette

And the boundary that doesn't move in either phase. No care records. No resident charts. No family details. No individual HR cases or performance records. If a document you're tempted to use has a person's name and circumstances in it, it isn't ready, and the answer is to find a different document, or make a synthetic one.

Chapter 4

Alex — AI host

Let's walk the exercise. It's the 90-day improvement planner, at careeconomycollective dot org, slash tools, slash improvement, then the word plan, the two words joined in the web address. It's free. The page says your figures stay in your browser, that the public tool doesn't send inputs to AI or save them on a server, and that the examples are synthetic. And it opens with a warning we like: targets are hypotheses until measured.

Eric Marquette

It has six parts. First, aim and workflow boundary. Write one sentence on what you're testing and where it stops. Our synthetic version: record a dietary preference change once, and have the kitchen sheet and the handover note draw from that single entry, in one unit, for one trial. Second, baseline and target. Enter the measured starting point, and a target you label as a hypothesis. If you hope for a reduction in double handling, write that as a hope, not a forecast.

Alex — AI host

Third, the change to test. One change. If you change the form, the tool, the handover and the meeting rhythm together, you won't know what did what. Fourth, balanced measures, and this is the one I'd spend the most time on. Time and cost is one dimension. Next to it, staff burden: double handling, frustration, overtime. Resident and family experience: responsiveness, continuity, how well things are communicated. Safety: near misses, missed handovers, dietary errors, compliance flags. And unmet need.

Eric Marquette

Unmet need is the one people forget. If the experiment releases some minutes, where do they go? Into a backlog of paperwork nobody had time for? Or into the walk, the conversation, the follow up call that used to slip? That's an honest thing to track, and it ties back to episode two's rule, that released capacity isn't cash. And on cost, count the full change cost: training, backfill, integration, subscriptions, governance, monitoring, and the protected improvement hours. Then the net picture might look thinner than the time saving alone, and that's useful to know.

Alex — AI host

Fifth, sponsor and review routine. Name the sponsor, the person with authority to unblock things, and the named owner, and fix the weekly review slot. Sixth, stop criteria. Write them before you start, when nobody is invested yet. Eric, give us a synthetic set.

Eric Marquette

Illustrative only. Stop if any adverse safety flag is linked to the new workflow. Stop if an extracted field is wrong against its source more often than the threshold you set. Stop if the manual fallback is used repeatedly, or fails when tested. Stop if staff rework is rising rather than falling, or if people have started keeping a parallel notebook, because that's a sign the new way isn't trusted. And the threshold numbers are yours to set. We won't invent them.

Alex — AI host

Which brings up who can pull the cord. Frontline staff, the named owner, the safety lead, and the sponsor should each be able to halt the trial immediately, without needing a meeting and without being blamed for it. If only the most senior person can stop it, you've built a trial that's hard to stop. Psychological safety here is practical: a person who sees a dietary error should be able to say stop in the moment.

Eric Marquette

And what could fail? A few things we'd watch for. Extractions that are wrong or drift over time, which is why the preview and human check exist. Staff building shadow workarounds. Alerts so frequent that people tune them out. A fallback nobody practised. And the quiet one, where the burden doesn't fall but relocates, so verified closure has to include asking the people downstream.

Alex — AI host

At 30, 60 and 90 days, the team makes one of three honest decisions. Stop, and end it cleanly, with no stigma, because learning that something doesn't work here is a result. Adapt, meaning change one boundary or parameter and test again. Or replicate, which we'd reserve for when synthetic and shadow stages have shown stable behaviour, no safety defects, and staff burden that's neutral or better, plus a cost picture that holds up with everything counted.

Eric Marquette

And expansion shouldn't be a single leap either. Even after a good result in one unit, the next unit is another bounded experiment, with its own baseline and its own named owner.

Alex — AI host

There's an objection I'd expect from a busy operator: this is a lot of process for a six minute task. Fair. If the whole thing costs more attention than the problem is worth, then stopping early is the right call. The planner is meant to make that visible, not to justify itself.

Alex — AI host

Let's close with the limits. Everything today was a synthetic design. The Collective has not demonstrated a provider trial or measured benefits. Any time saving remains a hypothesis until measured locally. We welcome products and ideas from every contributor as we work together on a care economy system for Australia. Contributors keep their existing intellectual property, and confidential details should only be shared under agreed protections. We are not giving legal, privacy or funding advice, and we have not called anything compliant or certified.

Alex — AI host

The one thing to take away: capture once, use well is a hypothesis, so test it small. One change, one owner, one date, a fallback that works, measures that include safety, staff, residents and families, and unmet need, and a stop rule agreed in advance. Then make an honest choice to stop, adapt or replicate. That's the end of the series. Thanks for listening.