A single-site home in Adelaide, a couple of years into AI

St Anna's Aged Care in Brompton is a stand-alone home. There is no corporate office behind it and no IT department, and quality sits with a couple of people rather than a department. For a couple of years now it has been putting AI into its clinical and compliance work, alongside a Sydney company called Innovation Philosophy. Almost everything it has done is within reach of a home of a similar size, and much of it started with someone asking the system a question and seeing what came back.

I visited on a day when Manos Katris was down from Sydney working through the details with the team. He built the system for Innovation Philosophy. At St Anna's I spoke with Amanda Birkin, the CEO, and Athin Christou, who leads the work there as Wellness, Allied Health and Technology Manager. Between the three of them I got a reasonable picture of how it all fits together.

What follows is what they are doing with it, how they set it up safely, and what has been worth the effort.

Where things stand

St Anna's was one of the first aged care homes Innovation Philosophy worked with. None of it arrived at once. It has gone in a piece at a time, which is how they wanted it.

The part they call clinical intelligence is live and being rolled out across the home. That is the day-to-day side: asking questions of their own records, drafting, trending, looking for gaps. Clinical governance, the oversight and reporting layer that sits above it, is built and close to being switched on.

Why a small home went looking

Larger groups have corporate offices behind them, with HR, quality and IT departments. A single-site home carries the same obligations with none of that support, and the conclusion at St Anna's was that they had to do it smarter.

Two things pushed them. One was time. An extra layer of compliance arrived at the same time as everything moved digital, so staff were spending more of the day working through systems and less of it on the floor.

The other was their own data. There was plenty of it and not much being done with it. Collating it meant spreadsheets and hours of work, and that process reliably surfaced the obvious things and lost the rest.

What it is being used for

Compliance and the new standards. This is the largest use and the one most homes will recognise. The system is trained on the standards and connected to the home's own documentation, so it can be asked where the gaps are against a particular standard, what evidence is missing, and what the home is not yet capturing.

The example that best shows what this changes: the lifestyle assessment forms were loaded in alongside the new standards, with a question about where the home was falling short, what to ask residents, and what to do with the answers. What came back was a table of gaps, one of which had not been spotted internally. They were capturing a lot about residents' pasts and much less about their present. That went into a team meeting as a set of real questions.

That was not work that used to take longer. It was work that never used to happen at all.

The related point is that the standards give overarching principles but not real-life examples, and a good deal of the value is in getting from one to the other.

Time given back at the desk is time available on the floor.

Looking further back after an incident. When a resident falls, the root cause analysis usually draws on the last week of progress notes, or the last month. Going back three or six months is possible but rarely happens. Something from three months ago may well be contributing to what happened today, and there is not much chance of finding it by hand on a busy shift.

There is a second effect here, on the quality of the work rather than the speed. A nurse asked to complete a root cause analysis in a hurry will complete it, but as a task to get done rather than a question to answer. By the time you have dug for three hours, you cannot really be bothered looking properly at what you found. Having the deeper information already sitting on the surface means the thinking happens while you are still fresh.

Trending feedback and complaints. Pulling three months of feedback together used to take about four hours before any analysis started. The system does that first layer of gathering and trending, and the team picks it up from there.

Writing and documentation. This came up more than anything else, and it was not what I expected to be writing about. Several staff at St Anna's have English as a second language, and the way it was put to me is that it is not what they know, it is what they are able to articulate. A nurse can hold a clear picture of a resident and still write a note that does not show it.

It is not what they know. It is what they are able to articulate.

The same holds for people who find writing hard for other reasons. One of the positive use cases has been supporting someone with dyslexia, who can have the conversation easily but stalls at the keyboard, spending longer working out structure than content. They now write the important points as bullets and let the tool assemble them. The intelligence is still coming from the person. The tool helps with the ordering and the delivery.

At St Anna's they compare it to the red line that appears under a misspelt word. It gives people confidence, because most people are smarter than what they manage to write down. There is also a view that it might keep people in the workforce longer, because the tiring parts get smaller and what is left is the part worth doing.

Lifestyle assessments were the first place the gap analysis was pointed.

How the system is set up

This is the part your board and your security people will ask about, and the answer is a good one. Innovation Philosophy does not host client data. The system sits inside the home's own tenant, next to its email, SharePoint and Azure, behind its own firewall and running on its own security standards. The model itself runs inside that boundary, so nothing leaves the organisation. The home's security officer signs it off before anything goes ahead. Innovation Philosophy's developer has worked as a CIO and a CISO, and treats keeping the security people comfortable as the first job rather than the last.

Nothing had to be replaced to make room for it. The systems St Anna's already run stay where they are, and the AI sits over the top of them as a layer that can read across all of them at once. At St Anna's that means their clinical software, Ausmed for training, the Alexys nurse call system and the roster. That is what makes the answers specific rather than generic. The example used to explain it is asking who worked in a particular wing yesterday, then asking what training those staff are missing given the notes and comments from that wing over the last month. You could do it by hand, but it takes long enough that nobody does.

The chat is one module. A second holds agreed reports, such as a quarterly resident analysis, which run at the press of a button because the work of defining and training them has already been done. A third reads the policy library, so staff can ask a plain question, like what to wear working in the kitchen, instead of searching 400 documents.

How the two organisations work together

Training happens in live sessions, with someone sitting next to staff and showing them how to ask questions. Everyone starts the same way: keep working as you normally would, run the AI alongside you, and when you finish, look at what it found and compare. Once someone sees it catch something they missed, they start trusting it. The human stays in the loop throughout, and it is never the first pass.

Before anything is switched on there are three stages: agreeing what the organisation actually wants and what people are worried about, checking whether the existing systems can support it, and then change management. On pace, the advice is that taking it all on at once is like drinking from a fire hydrant, and it is better to start with small things and small wins. The modules themselves get changed and improved regularly.

The time it gives back

Innovation Philosophy's figures are that a compliance response that would take one or two people five days comes down to around half an hour: roughly twenty minutes to assemble the evidence, and another five to seven for a first draft. Complex matters that would run to ten or fifteen days by hand come down accordingly. They are the first to say the numbers sound unlikely until you watch it happen.

There is a smaller example that says more about what the system is doing. A team had spent five days assembling a compliance response by hand. Running over the same records, the system surfaced something none of them had picked up: one of the residents had a dog that came to visit him, and that it was making a real difference to him.

Two things to expect along the way

Other software vendors may not hand over your data. Contractually the data belongs to the organisation and the product belongs to the software provider, but asking a provider for a replica of that data in the organisation's own warehouse has met real pushback. The estimate at St Anna's is that they would be two years further ahead if this had been sorted out earlier. The sector talks constantly about interoperability, My Health Record and FHIR as a common language, and this is where it gets stuck in practice. It is worth asking the question of your vendors early, and worth all of us pushing on it.

Some people will need time, and most of them come round. Innovation Philosophy sees resistance in most of the homes it works with, and the reason is usually the same one: people are worried about their jobs. Their experience is that the strongest objectors often end up the strongest supporters once they have used it. The bigger hurdle is imagination. People are handed something that will do almost anything and have no idea what to ask it, which is not far off sitting in front of a computer for the first time. That is what the live training is for.

Making it safe rather than making it secret

Staff will use AI on their own phones and there is no realistic way to stop them, so the question at St Anna's became how to make it safe instead. For a while, saying you had used AI felt like admitting to something you should not have done, which helped nobody.

The comparison they make is with the calculator. Nurses were once not allowed to use one for drug calculations, and calculators are now in everyone's pocket. The answer was never to ban them. It was to teach the principle behind the calculation and then let people use the tool. A governed system inside the organisation is a great deal safer than staff pasting resident details into a public chatbot.

Other things they have tried

A service robot at St Anna's, in front of the cafe.

AI is not the only thing St Anna's have brought in. Cleaning robots cover more than 800 square metres of corridor and communal space, which helps with infection control, takes manual handling risk off the cleaners, and gives them more time to spend talking with residents. There are smaller uses too, like posters and activity reminders. It is a home that is willing to try things, and that habit is probably worth more than any single piece of technology.