Perspectives

The layer nobody designed

István Borbíró · [publish date] · 4 min read

Every cancer team has one. The nurse who knows which surgeon answers messages and which one needs a corridor conversation. The coordinator who chases the pathology report before Thursday's meeting because it went missing before. The registrar who checks that the plan agreed at the tumour board actually reached the ward.

Ask a hospital where this work fits in the organisational chart and you get silence. It lives in people. It has no job description, no budget line, no metric. And it is the reason the plan survives contact with the week.

This layer has a name. Coordination.

And it has a history worth stating plainly: nobody designed it. Each protocol, each new therapy line, each new system arrived specifying the clinical act. Blood tests weekly. Imaging at week six. Review at progression. Who chases the blood tests, who notices the imaging slot got moved out another week, who carries the result to the person who can act on it: perhaps you know them, but it is not in their job description. The team filled the gap by hand, and the way they filled it hardened into how things work.

This is not only a healthcare observation. Natty Gur, an enterprise architect, reached the same conclusion from systems logic: "People were the integration, run by hand. You're not breaking the system, you're revealing it was never joined."

Coordination is a capacity. It can be assessed. It can be developed.

We do not leave clinical skills to chance; we train them, examine them, revalidate them. Coordination gets none of that treatment today. When it appears in policy at all, it appears as one condition on a list, next to funding and infrastructure and workforce. That placement misreads it. Coordination is the infrastructure layer that makes the other conditions load-bearing. Funding buys a tool. Workforce staffs a clinic. Whether the tool, the clinic and the pathway cohere into care is decided somewhere else. Sometimes in the corridor, case by case.

Then AI arrived, promising saved capacity. The quiet assumption is that the capacity a tool saves returns to the team. It rarely does, because the tool draws on the coordination layer it landed in. Someone curates what the model sees before the discussion. Someone owns the alert when it fires. Someone needs the time and the standing to question a confident output. Human-in-the-loop is a claim about the team. Corina Richter named the mechanism in an exchange about AI scribes: the capacity a tool saves flows straight into coordination work the system never measured. Context is not a condition the tool runs in. It is a resource the tool spends.

An honest objection deserves an answer here. Novel therapies stressed this layer for thirty years without producing a reckoning. Immunotherapy alone redrew who needed to talk to whom. Why would AI be different? Because a protocol makes demands on the coordination layer, and a tool makes decisions inside it. A protocol handed the team a list of acts and left the joining to them. The team absorbed it, invisibly, at cost. A tool operates inside the joining itself. The condition of the layer now shows up directly in what the tool produces. AI amplifies the coordination it meets. The research literature has reached the same seam from the individual side: Klein and Klein (2025) name the team level of AI's cognitive effects and stop there. The territory is open.

Meanwhile the layer is still held by hand. The people holding it are praised, depended on, and absent from every plan. Something that looks like infrastructure while it is still goodwill with a title. When one of them leaves, what disappears is the unseen working mechanism. Athena Hall, a physician writing on exactly this, put the design principle in one line: "Infrastructure isn't a person, it's a way of organizing care."

Getting there starts with visibility. Nobody funds what nobody sees.

What stays invisible stays personal: absorbed by individuals, carried into evenings, recorded nowhere except in how tired people are. Our work is making this layer visible and measurable, because a capacity you can see is a capacity you can build. The people who keep cancer care joined have been doing infrastructure work without the name. The name changes what can happen next.

Reference: Klein CR, Klein R. Frontiers in Artificial Intelligence. 2025;8:1719019.