Adult social care in England is carrying around 111,000 vacant posts and a turnover rate of 23.1 per cent, nearly three times that of the wider economy. As the over-65 population grows from 11.5 to 14.5 million by 2040, the sector will need an additional 470,000 posts to keep pace. Against that backdrop, the promise of artificial intelligence is easy to reach for and easy to oversell, so it is worth being honest about what the evidence actually shows.

The track record for AI in general is sobering. In a study of more than 300 enterprise AI deployments, MIT found that roughly 95 per cent delivered no measurable business impact. That number gets quoted as a verdict on the technology, but the more useful finding sits underneath it. The projects that worked had something in common. Organisations that partnered with a specialist were about twice as likely to see a return as those that built alone, and the biggest gains came not from flashy customer-facing tools but from taking friction out of routine back-office work. Technology was rarely the problem. How it was fitted to real work was. It is also worth noting that that study was in August 2025. That’s almost a year old, which, when it comes to AI, is a lifetime. The capabilities now are vastly different from what they were then. But even last year’s AI is clearly delivering tangible benefits for some.

What the general research says

In health and care the clearest wins so far have come from documentation. Ambient AI scribes, which draft a clinical note from a consultation, have been linked to real reductions in administrative load and burnout. Mass General Brigham (USA) reported a 21 per cent fall in burnout prevalence after a few months of using ambient documentation. These are promising results, but, again, they are early and are reflective of what the AI used to be like. It is extremely important when you engage with AI, to consider how fast the technology is evolving. We are accustomed to yearly upgrades. In AI, this happens monthly and sometimes weekly, and you need to be prepared for that in order to get the best out of it. If you are using AI primarily for automation of note-taking, That’s a good start, but you are behind in terms of what the technology is now capable of.

What we have seen in Australia

At Minikai, First and foremost, we care about the quality of care (Minikai’s founders have family members in long-term care), so we ruminated on the question: “What are the tangible benefits we are seeing now, and can we prove that AI is improving the quality of care?”

Take a nurse in a residential home writing up a fall at two in the morning. The next nurse on shift reads that note to know what to watch for, the GP relies on it, the family is updated from it, and an auditor later reads it as the evidence that care was delivered appropriately and in line with guidelines. When we looked at notes from nurses in an Australian aged care home, comparing heavy users of our system with light users over six months, the notes written with AI assistance were longer and more complete against the SBAR structure nurses use for clinical handover, and both gains were statistically significant. The nurse still decides what matters and what to escalate. The system helps get it onto the page, and as the provider put it, their nurses started finishing their shifts on time. The AI is also prompting the nurses and asking if they have documented everything correctly or sent reports that are mandatory. This is proactive training delivered at the point of care in the right context.

The same shows up in allied health. Minikai’s founding data scientist Xavier Gunn analysed 6,224 real interactions by 143 therapists across disability providers. The work almost always begins with retrieval, pulling together a person’s history, prior assessments and recent notes, and only then moves to the report, email or funding justification that follows. Across the twenty most common tasks, that returned more than 3,200 hours to clinicians. Mapped against the recognised tasks of each therapy role, AI assisted with between 23 and 55 per cent of the job. Across our customers in Australia and New Zealand, the same effect has returned more than 10,000 hours to care, with the heaviest clinical users saving upwards of 30 hours a month. This is happening now, but it’s just the beginning, and the frontier AI models improve significantly month on month.

The reason this works is not the model. It is that the information was captured for a purpose in the first place. A care plan exists so someone receives the right support, a funding assessment exists so resources are allocated properly, and a note exists so the next person knows what happened. When AI surfaces that context at the moment a clinician needs it, the work it was written for finally gets done, instead of the record sitting in a system nobody has time to read.

The deciding factor is the partner

My view, after two years of doing this alongside more than 25 providers supporting over 400,000 people, is that AI is transforming aged care, but only if it is done right. That means absorbing the administrative load and not the clinical judgement, keeping a person in charge of every decision that affects someone’s care, and building it with the clinicians who have to live with the result. The five per cent that work are not lucky. It comes from choosing the problem carefully and choosing the right partner to solve it. What I’ve seen from my 30 years of working in tech is that the best companies move almost as quickly as the frontier models do. I’ve seen that native AI companies (companies that started after AI became prevalent), move orders of magnitude faster than legacy software companies. Partnering with one of these AI-native companies is key if you want to stay at the cutting edge of the technology while also managing it safely.

The technology is ready. The deciding factor is who you trust to deliver it together.

Luke Janssen is the Executive Chairman of Minikai, a company that builds person-centred AI for disability and aged care across Australia, New Zealand and the United Kingdom. He is happy to talk through what this looks like in practice for a UK provider, including live examples from organisations like yours: luke@minikai.com