The last time I read a 200-page document word for word

The last time I read a 200-page document word for word was probably before I learned how to adapt. In a recent meeting, I was asked if I had read a project description that was over 150 pages long. When I answered ‘no’, the room filled with a mixture of consternation and disbelief. I let the silence hang for a moment before adding, “I haven’t read it, but I know what’s in it”. The raised eyebrows told me I had their attention.

What they didn’t know is that one of my personal processes has fundamentally changed. Instead of manually reading long documents, I now drop them into a secure chatbot, ask for a summary, and then ask targeted questions. Not only do I grasp the contents much faster, but I also retain the information for longer. This is a small, personal evolutionary adaptation. It’s a response to an environment that is changing at a dizzying pace. Now, imagine scaling this kind of adaptation across an entire healthcare organisation.

The Cambrian explosion of AI

The current generative AI innovation cycle is arguably the fastest in technological history. It reminds me of the early 2000s smartphone boom, when a new model with a significantly better camera appeared every few months. But today’s cycles are far more compressed and operate on multiple layers simultaneously. The half-life of AI model relevance is now shorter than the average hospital committee meeting.

We are witnessing a multi-faceted evolution:

Foundational capability leaps: These are the big, paradigm-shifting changes, like the move from text-only models to true multimodality, where AI can see, hear, and speak. These major leaps have been happening on a 12 to 18-month cycle.
State-of-the-art performance jumps: This is the equivalent of the smartphone megapixel war. It is not a new capability, but a dramatic improvement in an existing one. A new model that significantly pushes performance forward arrives every three to six months from a major lab like OpenAI, Google, or Anthropic.
Ecosystem and democratisation velocity: This is the speed at which cutting-edge capabilities, once exclusive to expensive, proprietary models, become available in smaller, faster, open-source alternatives. This is happening on a weekly to monthly basis, with the gap between closed-model capability and its open-source replication shrinking dramatically.

These cycles feed into each other, creating a relentless, accelerating cascade of innovation. The ground is constantly shifting beneath our feet.

Building an organisational immune system

So how does a healthcare organisation, with its expensive legacy systems, strict regulatory frameworks, and deeply entrenched clinical workflows, innovate in such an environment? You don’t build a rigid structure bolted to the ground; that would be foolish.

Instead, you must think of your organisation like a biological organism. It needs an immune system. An effective immune system doesn’t rebuild the body every time it encounters a new virus. It develops adaptive antibodies that can identify and respond to new threats and opportunities without disrupting core functions. For healthcare, this means building an AI strategy that works around, not through, the fragile ‘glass houses’ of our existing electronic health record (EHR) systems.

This can be achieved with a practical, three-tiered approach that builds resilience and capability over time.

Tier 1: Low-risk administrative tasks. This is the first line of defence. It involves deploying AI for tasks that do not directly involve patient care. Think of AI-powered scribing tools that listen to doctor-patient conversations and automatically draft clinical notes, freeing up clinicians from hours of paperwork. Or automating prior authorisations to speed up billing and reduce denials. These applications introduce AI safely, delivering immediate value and building organisational comfort with the technology.
Tier 2: The smart layer. This tier involves creating intelligent tools that sit on top of existing systems without altering the core records. These tools can, for example, instantly summarise a patient’s entire medical history for a clinician or translate complex discharge instructions into simple language for a patient. They read from the EHR but do not write to it, providing powerful support while respecting the integrity of the core system.
Tier 3: Deep clinical integration. This is the future, where AI is woven into the fabric of clinical care. This includes AI that assists in diagnosing diseases from medical scans or helps create personalised treatment plans based on a patient’s unique data. For these highly sensitive applications, trust, control, and customisation are paramount. This is where open-source models become the killer feature, allowing adaptable, distributed, secure environments while being fine-tuned on its specific protocols and patient populations.

Your mutation lab for the future

To prepare for this third tier, you cannot wait. You must begin building the capability now. This requires creating a small, dedicated ‘AI skunkworks’ team, a kind of mutation lab for your organisation. These are your digital insurgents, empowered to experiment and adapt. Their goal is not to launch a product tomorrow, but to build the internal expertise so that when the time is right, you control your own destiny.

How do you get this mutation lab started? By focusing on small, rapid, and manageable actions, for example:

The 90-day experiment
The three-person rule
The €5,000 proof of concept
The sandbox principle

This team will grapple with the critical questions: How do we convince clinicians to trust AI when the models change faster than medical guidelines? What happens when our team develops capabilities that challenge existing workflows? By tackling these dilemmas on a small scale, you build the cultural and technical antibodies needed for a large-scale transformation.

The new paradigm is a shift away from monolithic systems and towards an open, agentic ecosystem where AI can dynamically interpret clinical data. The future belongs to organisations that treat this transformation as a continuous evolution. Those who start rewiring for decision velocity and adaptability today will lead tomorrow’s AI-powered healthcare landscape.

My personal PDF reading habit is a tiny adaptation. But it proves a vital point. When you start evolving your own processes, you begin to see the immense potential. You stop seeing AI as a threat to be managed and start seeing it as a force to be harnessed, one small, brilliant adaptation at a time.

💥 May this inspire you to advance healthcare beyond its current state of excellence.