Every visit to a GP results in data (think blood pressure, age, risks, symptoms). This data means the healthcare sector is primed for AI, which could analyse results against patient histories and medical journals to rapidly deliver the most accurate diagnosis.
But the sector is in an AI holding pattern, as there’s a need to strike a delicate balance between deploying AI and ensuring data privacy and governance. In one prominent example of how delicate that balance can be, an AI tool was developed in Australia which can detect up to 124 chest X-ray findings in under a minute, flagging possible findings and triaging the most critical cases so radiologists review those first, leading to quicker diagnoses and patient treatments. But while the images used to train the model had been sufficiently de-identified, its development came under scrutiny as data and imagery were used reportedly without the consent of patients.
This is the scrutiny the sector is under and, on the surface of it, that scrutiny is understandable. But the risk of not using AI and the insights it can glean, I would argue, is riskier than not using it at all.
So… what seems to be the problem?
This is not to say healthcare organisations aren’t using AI in some form. For one, 22 per cent of Australian GPs are using AI as a clinical scribe to automatically generate notes, care plans, and orders for tests following patient consultations.
Recent research found Australian healthcare and life sciences organisations rank second only to advertising and media in their overall uptake of generative AI and large language models, with 47 per cent report leveraging these technologies across many use cases, compared to 38 per cent in all other sectors.
But a deeper dive into the data paints a more complex picture with regards to how AI is actually being deployed within healthcare organisations. That same research also revealed that adoption lags peer industries in one very critical area; customer service.
Healthcare organisations are significantly less likely than their peers to use generative AI to proactively identify customer issues (39 per cent versus 49 per cent across other sectors) or to guide patients toward the right self-help resources (47 per cent versus 60 per cent).
It’s this gap between broad awareness and deep, operational integration where healthcare’s AI opportunity is being left quietly in the waiting room. Patients today interact with their healthcare providers across a growing number of digital touchpoints, from scheduling appointments and navigating benefits, to managing chronic conditions and understanding treatment options. When those interactions are frustrating, confusing or slow, the consequences could lead to delayed care, eroded trust and poorer outcomes.
Because if you look at what AI can provide clinicians, it’s insight that, in the absence of AI, wouldn’t exist otherwise. From there, the clinician can analyse that insight, act on it, question it or ignore it – but the more information a clinician has, the more perspectives they can use on to inform a diagnosis.
Put another way: wouldn’t you rather have an insight than not?
We need to walk and chew gum at the same time
Australia has been through the Medicare breach and, more recently, the theft of sensitive medical information including the medical histories of patients and donors from IVF provider Genea Fertility.
Therefore, an understandable root cause of the sector’s apparent hesitancy towards AI could be one vital aspect: the regulatory and privacy dimensions of AI deployment. However, research found that healthcare organisations are only marginally more concerned about the regulatory implications of agentic AI than their peers in other sectors (25 per cent versus 22 per cent).
This sounds like a laissez-faire approach to privacy but, given the benefits AI can bring, is the right posture. And additionally, when it comes to things like governance, privacy and more, so long as you own the data, it’s yours: if it’s in an AI data cloud — rather than a commercial, consumer-focused public product – those things should be assured.
In short, bring the AI to the data, not the other way around. Don’t send the data out to external APIs, for example. And with encryption, role-based access controls, dynamic data masking, and governed data sharing (which lets organisations collaborate on live datasets without surrendering control of the underlying information), the path to AI can be significantly shortened.
Though it’s worth noting that not all AI is the same. For those that differ, implementing natural language processing to SQL capabilities or implementing AI functions into engineering workflows can be tested and validated for safety quickly, and approved for deployment.
Importantly, when it comes to AI in healthcare, we need to walk and chew gum at the same time: by that I mean build a robust data strategy and an AI strategy concurrently. We can spend all the time in the world building one thing ahead of another, but in that time, we’re not getting the insights AI can provide.
The benefit of insight shows why healthcare must catch up
While healthcare organisations are certainly on the AI journey, they need to – and can – do more.
In any other workplace, if you’re given the option of being provided an insight to consider versus an absence of insight, you’d take the insight every time. This should not be any different in healthcare – an insight can be analysed, operationalised, questioned or even ignored… but not having it in the first place reduces the information available and the potential benefit to patient outcomes.
AI can provide unique insights at a speed once thought unfathomable, and healthcare organisations need to shift their mindsets from AI trepidation to AI adoption.
Because it’s not hyperbole to say doing so could save lives.
Glenn McPherson is the Regional Vice President, Australia, for AI Data Cloud Company Snowflake.
Add Comment