The conversation almost always starts the same way. A CEO or quality lead leans forward and tells me they know they need AI. They have been to the briefings, read the reports, watched a colleague's organisation begin something they don't fully understand yet. What they don't have, they say, is a clear sense of where to begin in their own.
This is not really a knowledge gap, it is an organisational one, and the providers who navigate it well are the ones who refuse to treat it as a technology decision.
The real barrier is not technology, it is clarity
When we work with providers across NDIS and aged care, the blockers are almost never technical, they are organisational. The team is not sure which problems AI can actually solve for them. The board is not sure what good looks like. There is a reasonable, healthy caution about introducing new tools into environments where someone's safety and dignity are on the line.
That caution is a feature of the sector, and care providers ought to be careful about this. The risk worth worrying about is the one that arrives when thoughtfulness tips into inertia and the conversation about AI quietly stops moving. The providers who avoid that outcome do not do it by importing someone else's roadmap; they do it by getting clearer about their own work, and by choosing a starting point anchored in their own reality.
Start with the work, not the technology
The most common pattern I see is an organisation starting with a tool and trying to find a use for it, and that gets the order wrong. Start with the work itself, the practical reality of what your team does on a Tuesday, and ask where time is being lost to admin that could be spent with people.
In most care organisations, the answer is sitting in plain sight. Support workers spend large portions of every shift on documentation, including progress notes, incident reports, handovers, and compliance evidence. Quality teams lose entire days assembling audit packs. Coordinators flip between half a dozen systems trying to assemble a coherent picture of the week.
The actual problem AI in care should be solving
Before talking about where to begin, it is worth being precise about what we are solving. The pattern across the providers we work with is consistent. Data is fragmented across multiple systems, which makes a holistic view of any one resident or participant harder than it should be. The work is heavily regulated, which means a meaningful share of every shift goes into producing compliance evidence. And the financial pressure is now endemic, particularly as sector reforms reshape how funding flows.
Those three things, fragmented data, the weight of compliance, and financial sustainability, are the pressures AI in care should be attached to. The work is not to add AI for its own sake; the work is to anchor a specific use case to one of those pressures.
That is also why documentation is the right place to start. Documentation exists so the next person picking up that person's care knows what happened, and audit evidence exists so the people receiving support can be confident in the quality of that support. Somewhere along the way, those things stopped being for the person and started being for the system, and AI used well returns time to the work it was meant to support.
A practical place to start
The starting point I most often suggest is a one-off pass of the quality and audit reports your organisation already has to produce. Run it over the documentation you have today, and let the AI interrogate what is there. This surfaces two things at the same time: the places where there are quality or compliance issues, and the places where funding is not being captured properly.
The objection I usually hear next is about data quality, often phrased as “our documentation isn't clean enough, we're not data-ready”. The way I think about it is that the gap is the insight, because the places where the AI cannot ground a response are exactly the places your team needs to know about. The gaps tell a story, and the story is usually about where compliance is at risk and where the next conversation needs to happen.
From that view, the next step opens up. The same platform that helped you see the gaps can prompt the people capturing information to close them at the moment of capture, so documentation improves as it goes in. And once you know what quality you want embedded in practice, introducing AI as a clinical support tool to frontline staff becomes much more straightforward.
The team is the project
Getting the technology right is one part of the work, but bringing your people along is the other, and the second matters more.
Care workers have heard a lot of promises about technology that was going to make their lives easier, and frankly, a meaningful proportion of those promises did not land. The most useful thing you can do is be honest about what AI will and will not do, and then demonstrate something specific. The early sequence we usually run, audit reports first, then better capture at the point of work, then frontline support, is also a sequence the team can follow with confidence, because each step earns the next one.
Why a purpose-built platform matters
The general-purpose AI tools on the market are remarkable, and they will keep improving. The question is not whether the underlying model is good, it is what the platform around the model knows about the work it is being asked to do. Care organisations are not general-purpose environments. The language is specialised, the regulations are specific, and the consequences of getting it wrong, for the person receiving care, are real.
This is why we built Minikai for the care and support economy, and why the design approach we describe is person-centred AI. Each of our agents, the ones we call Minis, is centred around one person, a participant or a resident, and holds the context relevant to that individual. That structure is deliberate, because care happens one person at a time, and the technology should reflect how the work actually flows. The Mini understands the regulatory environment the team operates in, the way notes need to be structured, and the language clinicians and support workers actually use.
Governance, privacy, and auditability are not features bolted on at the end; they are foundational, because in this sector they always have been. When a Mini drafts a note or surfaces information about a participant, there is a record of its provenance: what it produced, what informed it, and who approved it.
The benefits compound
The early value of AI in care is the obvious one: less time spent on admin, more time spent with people. What I find more interesting is what happens after that, because the value is cumulative.
When evidence is structured at the moment it is captured, compliance teams stop having to assemble it later. When the context for a person is held coherently rather than scattered across systems, coordinators get real operational visibility, anchored in what is actually happening across services. And when that context follows the person between providers, shifts, and clinicians, families experience continuity of care that the sector has historically struggled to deliver.
Over a longer horizon, what emerges is an evidence base that reflects how care is actually delivered, the ability to identify changing needs earlier, and more personalised support planning at scale.
Where I would start
If you are still not sure where to begin, you do not need a transformation strategy or a five-year plan, you need three short conversations.
Talk to your support workers about where they lose time. Talk to your compliance team about what keeps them up at night. Talk to your leadership about what they would want to see in six months if this went well, and what they would consider a failure.
Then find a partner who understands the sector and can meet you where you are. The journey starts with one honest look at the work, and one step toward making it better.