Leading the Shift: Reimagining Clinical Practice in an AI-Enabled Future

---

Last week I closed the AI in Clinical Practice conference with a provocation I hoped the room needed to hear. While the day had showcased impressive AI deployments across New Zealand’s health sector, I argued that the most important question was still going unasked: not whether AI works, but whether we are applying it to the right problems.

The pattern we keep repeating

I started with a historical lens. When factories first adopted electric power in the late 1800s, they simply replaced steam engines with electric motors, keeping the same layout, same process, with only slight improvement. It took thirty years before anyone asked why the factory needed to be built around a central shaft at all. When they finally did, productivity was transformed.

Computing followed the same arc, understood first as faster calculation and only later recognised as an entirely new information environment. My point is direct: every major technology shift has been underestimated in exactly this way. People see the tool. They miss the system change underneath it.

"AI is familiar in pattern, but different in consequence."

What makes AI different this time is pace (ChatGPT reached 100 million users in two months; the telephone took 75 years), breadth (it operates in knowledge work and professional judgement, not just task execution), and irreversibility. Once embedded in clinical workflows, these choices will be hard to unwind.

The risk is underutilisation

Underutilisation isn't just low adoption or failed pilots, but something subtler and more costly: using extraordinary capability to improve inherited processes without first asking whether those processes deserve to exist in their current form.

The clinical letter is my centrepiece example. AI scribes can now produce excellent clinical letters, and that is genuinely valuable. But the letter only exists because there was historically no better way to transfer structured knowledge between practitioners in different settings. That constraint is disappearing. Optimising the letter, rather than questioning whether it should exist at all, means missing the deeper opportunity.

"The risk is not that AI fails in healthcare. The risk is that we succeed, at making the system only slightly better than it already is."

Take the fax machine, still the primary document transport mechanism in global healthcare, with around 70 percent of fax traffic health-related. We risk applying PhD-level analytical capability to writing better fax messages.

Augmentation, not replacement

As I've written before, the labour substitution narrative doesn't work in health. Healthcare’s problem is not excess labour but insufficient capacity. The right framing is augmentation: better outcomes with the clinicians we already have.

In practice, that means AI improving the starting conditions for clinical judgement rather than replacing it, reducing the administrative burden that drives burnout, and freeing experienced clinicians to spend more time on complex care. The family conversations, the navigation of ambiguity, the coaching of junior staff: these are what get crowded out by administrative drag.

Redesigning how knowledge flows

The bigger vision is a shift away from episodic, fragmented clinical knowledge toward something more continuous and structured. Today, a complex patient’s picture is scattered across the GP, the specialist, the ward, the community nurse, and the ED, and rarely fully reconciled in real time. AI makes it possible to synthesise those signals before the next consultation even begins.

That is a different architecture for care delivery, not just a better document. .

Leadership is the bottleneck

Technology adoption is easy. System change is leadership. I’ve seen three common failure modes after a pilot succeeds: adding technology to an existing workflow without removing anything; optimising one part of the system while creating friction elsewhere; and deploying the tool without doing the process improvement and cultural work that makes time savings real.

We should challenge the assumptions behind a process before approving the next pilot, design explicitly for augmentation, embed trust and governance early, and measure downstream outcomes rather than just user numbers.

"What is this process actually for? Ask it before approving the next pilot. Ask it before scaling the last one."

We know the historical pattern of underestimating transformative technology. We do not need to wait thirty years this time. The question for every organisation reading this is whether you will use AI to redesign how knowledge flows through your system, or whether you will still be optimising your letters when the real shift arrives.

---

CIO Studio provides independent digital strategy and leadership for New Zealand's health, NGO, and community organisations. If you want to talk to an expert about your digital strategy, get in touch for a no obligation conversation.