The NICU Nurse and the AI Algorithm

A few days ago I was talking to a hospital liaison named Julia about her work, and she shared with me an anecdote about her favorite job over the course of her career, a stint as a NICU nurse. Three years of her life, she told me — the hardest work she had ever done, and yet the most rewarding. Years later, she was reflecting how she still wonders about the babies she cared for and wants to find them. She hopes one of them is a teenager now, living some ordinary life, with no idea a stranger still thinks about them.

Julia's reflections tell a story about the intimate unspoken element of healthcare, about what it means for someone to need a doctor and to put one's life in another (random) person's hands.

At my work at Magicare, I work in what seems to be the hottest intersection of industries in the world, Healthcare & AI. Despite being classified as the tech team running data science, I somehow find myself talking to people at healthcare organizations almost every day. I hear the same theme echoed in Julia's story constantly, people looking for solutions to move away from the administrative work so they can do what they signed up to do - take care of people who need it. And I keep landing on the same question: as AI gets better and better taking over more and more of the space, what happens to the care?

Artificial Intelligence's whole sell is optimization. Remove friction, remove steps, remove the human bottleneck. That's a good instinct most of the time. But healthcare isn't called the health industry. It's called healthcare. The second half of the word is the part that's hard to automate, and the part that actually makes the experience good or bad.

The second half of the word "healthcare" is the part you can't automate, and the part that actually makes the experience good or bad.

Think about the difference between a doctor who delivers the hard news bluntly without emotion and one who delivers the same news with real sensitivity. It can be the same grim diagnosis. Yet a completely different experience. The caring isn't decoration on top of the medicine — it is a huge part of what patients are actually looking for when they walk through the door.

So where's the line? I recently saw a startup out of Utah, Nolla Health, announce it had gotten regulatory approval to prescribe medication with no doctor involved at all (albeit low stakes with strict protocol to prescribe acne medication). My first reaction was: wow that's cool, one less appointment & one less barrier to medicine. My second reaction was a more uneasy one. If I don't know where to start and I let an AI walk me straight to a diagnosis, is that progress? Or are we thinking about this all wrong?

I think about this at my own company too. Magicare's secret sauce is using top of the line AI to read through the long patient documentation PDFs that come with every referral, and pull out the data a nursing facility actually needs to decide if they're the right fit for that patient in seconds. Nobody wants to read a hundred pages of paperwork. It's tedious, it's slow, and it's exactly the kind of task where humans make careless mistakes. When we take that off someone's plate, we're not saying "do this with fewer people." We're saying: "go spend that time actually building the care plan. Actually caring for the patient."

While I strongly believe in Magicare's mission, I don't think my company is the hero here, and I don't think Nolla Health is the villain. I think they're both just data points in the same experiment the entire industry is running right now. We're still early in this. Nobody fully knows yet exactly what role AI will end up playing in healthcare, or how big that role will be.

But I think I know which way it has to go. AI should speed things up. It should clear away the boring, error-prone, soul-crushing parts of the job. What it shouldn't do is quietly replace the part of healthcare that was never the bottleneck in the first place, the part where someone looks at you and actually cares what happens to you. No (good) doctor or nurse should be concerned about losing their jobs to AI.

The companies that build AI to elevate humanity, not replace it, are going to win. The ones building toward replacing humans entirely are going to lose — because the thing patients are actually asking for was never strictly the information. It was always Julia, years later, still wondering about her precious NICU baby.