The challenge is to move faster while preserving the context, meaning and accountability humans have historically supplied.
I’ve experienced the consequences of the first part of my headline personally. And it’s why I’m increasingly concerned that, in our rush to solve it, we don’t create the second.
Something clicked for me over the weekend. We seem to be converging on a fairly binary debate about technology in healthcare.
On one side: interoperability has taken too long. We’ve talked about standards for years. Other industries solved this. Technology can do it. Incumbents need to open up and get out of the way.
On the other: healthcare is complicated. Standards take time. Safety matters. We need governance before we move.
Both contain some truth. Neither is enough.
Healthcare is genuinely more complex than banking. But we also have a habit of using that complexity to explain why things haven’t moved.
The more important distinction isn’t between moving fast and moving slowly. It’s between barriers created by legacy constraints or unnecessary restrictions on access to information, and foundations that exist because meaning, safety and accountability matter.
We need to remove the first faster, without bypassing the second.
For me, this isn’t an abstract argument.
In my own care, clinically important information wasn’t available when and where decisions were being made. The consequence wasn’t simply duplication, inconvenience or another appointment. It affected years of subsequent care and treatment.
Six years later, when the original information finally became visible through My Health Record, the significance of that earlier clinical pathway could be understood. By then, some consequences couldn’t be reversed.
I had undergone IVF treatment that, with that information available earlier, may have been approached differently. By the time the pieces were finally connected, alternative treatment pathways that may have changed my fertility outcome were no longer available to me.
I can’t know the counterfactual. I can’t say with certainty what would have happened if every clinician had had the right information at the right time. But I do know what it means when we describe poor interoperability as an inconvenience.
Sometimes the opportunity lost is irreversible.
That experience makes me profoundly impatient with information that cannot follow a patient. It also makes me profoundly uncomfortable with the idea that simply making all information flow is the answer.
Because information without its meaning can also harm.
Methotrexate is a confronting example.
In conditions including rheumatoid arthritis it is commonly taken weekly, and errors involving daily dosing have caused severe toxicity and deaths.
The medicine name alone isn’t enough. Indication, dose, frequency and clinical context are part of the information required to use it safely. Data can be available and still be unsafe if the meaning required to interpret it has been lost.
That is the tension I think we’re failing to articulate.
Banking is often used as the comparison. There is plenty healthcare can learn from it, but a dollar is still a dollar when it moves between banks.
Healthcare data doesn’t behave like that.
A diagnosis, medication or observation can mean different things depending on who recorded it, when, why, using which terminology and in what clinical context. The problem isn’t simply moving data – it is preserving enough meaning that the person or system receiving it can safely act on it.
Related
And this isn’t uniquely Australian. The same tension exists across health systems with very different infrastructure, regulation and approaches to health data.
That isn’t an excuse for poor interoperability. It’s why semantic interoperability, identity, provenance and clinical context matter.
AI has made this much harder to ignore.
Agents can navigate systems, assemble information and increasingly execute workflows that previously required a person. Where direct integrations don’t exist, emerging approaches can even interact with existing user interfaces.
That is extraordinary progress. But automating access to fragmented information does not make the information interoperable.
An agent may be able to read what is on a screen. That doesn’t inherently establish whether it is current, authoritative, complete, correctly attributed or clinically equivalent to information somewhere else. Once an agent can act rather than simply summarise, those distinctions become much more consequential.
This is where capability and governance answer different questions. Seeing a product work can answer “can it do this?”. It doesn’t necessarily answer “under what conditions should it do this?”
Take consent.
A clinician automating part of their workflow sounds straightforward. Now put that clinician across several practices, different clinical systems and different organisational governance.
An agent may read information from one system, combine it with another and initiate an action somewhere else. What exactly has the patient consented to? Which organisation authorised that use and remains accountable? What becomes the authoritative record?
These aren’t arguments against the technology, nor does raising them imply that a particular product hasn’t addressed them. They illustrate why architecture matters.
Once systems can move information across organisational boundaries and act on clinical records, technical design starts determining things clinical governance needs to control: authority, provenance, attribution and accountability.
Not every technical detail needs to be public. But the parts that determine clinical risk can no longer be treated as purely technical implementation details.
The same applies to standards.
Standards are sometimes positioned as the slow alternative to simply building. But good standards development isn’t waterfall. Establish enough common ground to build, implement it, test it against real workflows, find where it fails and iterate. That should sound familiar to anyone who has built a product.
Standards aren’t the finished building. They’re the shared agreement about where the foundations go so everyone doesn’t build a different staircase.
This is where I think AI creates a genuine inflection point.
For years, poor interoperability has meant duplicate entry, PDFs, phone calls and clinicians manually reconciling information.
Humans are effectively the interoperability layer.
But they do much more than reconcile data. They supply context, judgement and accountability. They know an old medication list isn’t necessarily current. They recognise when apparently conflicting information can both be valid. They understand why something was documented in a particular place. And when something doesn’t make sense, they stop and ask.
Much of that was never encoded because the system was designed around a human supplying it.
Agentic AI changes that assumption.
As AI takes on more of the work humans have historically done to reconcile healthcare information, we need to recognise that humans weren’t simply moving data between systems. They were supplying part of its meaning.
That makes better interoperability more urgent. It also makes semantic integrity, provenance, governance and accountability more important, not less.
So yes, we should move faster.
Closed ecosystems should be challenged. Core clinical information should be able to move safely with the patient, and the APIs that enable that shouldn’t be unnecessarily difficult to access. There is enormous value to build on top of that foundation. Access to information required for care shouldn’t itself become the barrier. But moving faster doesn’t require pretending every source of friction is the same.
Information that doesn’t move can harm. Information that moves without meaning can harm too.
The answer to the second cannot be to tolerate the first.
The challenge is to move faster while preserving the context, meaning and accountability humans have historically supplied.
AI is rapidly removing the technical question of whether we can automate healthcare.
The leadership question is whether we understand the information well enough to know what must remain true when we do.
Danielle Bancroft is the boss of product strategy (interoperability and integrations) at Telstra Health. She is the founder and managing director of Off Label Consulting.
This article was first published on Ms Bancroft’s LinkedIn feed. Read the original article here.