The Machine Didn't Wait: What Musk's Confession Means for Perfusion

Elon Musk was recently asked what happens to people when the machines no longer need them. He didn't soften it.

"There will be fewer and fewer jobs that a robot cannot do better. These are not things I wish would happen. They probably will."

Sit with that second sentence. He is not celebrating. He is not selling a vision. He is telling you what he believes is inevitable and admitting he wishes it weren't. That is not optimism. That is a confession.

I've been thinking about that confession in the context of our profession. Perfusion. A job most people outside cardiac surgery don't know exists, performed by a small, quiet group of clinicians who sit behind the surgeon and keep patients alive while their hearts are stopped. If you had to design a job that looked safe from automation, you might design this one. It's rare. It's high-stakes. It requires years of training. It involves judgment under uncertainty, a chaotic surgical field, and the legal weight of being the person responsible when things go wrong.

And yet I don't think we should kid ourselves.

What the machine can already do

A modern heart-lung machine is already, in large measure, a machine doing the work. We tend to forget that. The pump is automated. Gas mixing is automated. Anticoagulation monitoring is automated. Temperature control is automated. Level detection, bubble detection, pressure monitoring, all automated. The newer systems will wean onto bypass themsleves. The perfusionist sits at the console integrating those data streams, making judgments, intervening when the automation can't handle what's in front of it.

That last part matters. The perfusionist is there for the exceptions, not the rule. Most of the case runs itself, and has for decades. The craft sits in the margins. In the five minutes out of a four-hour case where something drifts, or something unexpected happens, or the patient's physiology refuses to behave the way the textbook said it would.

The honest question isn't whether automation is coming for perfusion. It's already here. The question is how much further it goes, and what's left for the human on the other side of it.

The case for evolution, not extinction

I want to be careful here because I think the honest answer sits between two comfortable lies.

The first comfortable lie is that perfusion is safe because it's specialised. That's what radiologists were told. It's what anaesthetists told themselves when J&J launched Sedasys, the automated sedation system that was going to replace anaesthesia assistants for colonoscopies. Sedasys got pulled from the market in 2016 after failing commercially, and people took that as proof that clinical automation had limits. It didn't prove that. It proved that the first attempt was premature, that the economics didn't work yet, and that hospitals weren't ready. Those are solvable problems. They get solved.

The second comfortable lie is that AI will replace us all by Tuesday. That's not true either. The tail of weird cases in cardiac surgery is long. Patients don't read the protocols. Circuits fail. Cannulas migrate. Surgeons do things that weren't on the plan. Someone has to hold accountability when the patient is on the table, and the regulatory system is nowhere near handing that accountability to software.

The realistic answer is somewhere less satisfying. Partial automation will continue to erode the routine parts of the work. Closed-loop systems will handle more of the in-case management. AI-assisted decision support will get better at flagging deterioration before the human sees it. The fraction of the case that requires active human input will shrink.

That doesn't mean perfusionists disappear. It means the job changes. The person at the pump becomes less of an operator and more of a supervisor. Fewer hands-on tasks. More interpretation. More accountability for outcomes across multiple cases running simultaneously. One perfusionist overseeing two theatres, then three, then more. We've seen this movie in other fields. It ends with fewer practitioners doing higher-leverage work, and a long painful period where the ones who didn't adapt found themselves on the wrong side of the transition.

The part Musk got right

What Musk got right, and what most commentary on AI and work misses, is the meaning question.

"How do people then have meaning? If there's not a need for your labor, what's the meaning? Do you feel useless?"

He said that was the harder problem. Not the economics. Not the policy. The question of what happens to a person who built their entire identity around being needed.

Perfusionists are particularly exposed to this. Not because we're more fragile than anyone else, but because of who the job attracts. The profession self-selects for people who want to matter in the room. People who take pride in being the one who notices the drift before anyone else does. People who've built their sense of self around the fact that the patient is alive because they were there. That's not vanity. That's a reasonable response to doing the work. But it also means that if the work thins out, if the machine takes the 80% that used to feel like the job, the people left are going to feel something they don't have good language for.

It's not unemployment. It's something worse. It's being in the room and not being necessary.

That's the thing I don't hear anyone in our field talking about, and I think we should. Not because it's coming next week, but because the generation of perfusionists training now will live through it, and the ones who thrive will be the ones who worked out early that their value can't be wrapped up entirely in the technical task.

What's actually defensible

If you strip the technical task away, if you assume for the sake of argument that the pump runs itself within a decade, what's left that a human perfusionist does that a machine can't?

A few things, I think, and they're not the things we usually emphasise in job descriptions.

The first is accountability. Someone has to be legally and morally responsible when a patient is on bypass. That person can't be a software vendor. The regulatory environment may evolve, but it doesn't evolve that fast, and there's a reason hospitals want a human name on the record. Accountability is expensive and humans are cheaper than liability insurance.

The second is judgment in genuinely novel situations. Not the common complications. Those get automated fast. The weird ones. The case where the cannula is in the wrong place and the surgeon can't see it and the pressures are doing something strange and you have to make a call with incomplete information in ninety seconds. AI will get better at this. It won't get there first.

The third is relational. Perfusion isn't a solo sport. It's theatre work. It involves reading the surgeon, coordinating with the anaesthetist, managing the tempo of the case. Automation is bad at this and will stay bad for a while, because the relational work is undefined and context-dependent and involves understanding what someone else is about to do before they do it.

The fourth is teaching, research, mission work, clinical development. The things we've always treated as ancillary to the real job. I'd argue they stop being ancillary. They become the job. The perfusionist of 2040 who is only a pump operator is going to have a bad time. The perfusionist who is also a clinical scientist, an educator, a program developer, a person with a body of work outside the theatre, that person is fine.

What I'm telling my own team

I run a perfusion services company. I employ perfusionists. I have a responsibility to think honestly about where this is going, and I'd rather do it in public than in private.

What I tell my team, and what I'd tell anyone entering the field now, is this:

Don't build your identity around the task. Build it around the judgment. The task is the part that gets automated first. The judgment is what's left.

Get good at the things that don't show up on a checklist. Teaching. Mentoring. Going to the places where no one else will go. I've taken cell salvage to the Solomon Islands, paediatric heart surgery to Fiji, Rwanda and PNG, and I'll tell you, nothing I did in that theatre was automatable, because the automation doesn't exist in Honiara and the nearest service engineer is a plane ride away. The work that's hardest to automate is the work that's hardest to standardise. Seek it out.

Pay attention to the business you're inside. Understand the economics of what you do. Know why your service exists, who pays for it, and what would happen if it stopped. The clinicians who survive transitions like this are the ones who understand the system well enough to see where they fit when the system changes shape.

And take the meaning question seriously now, before you have to. Work out what your life is going to be about if the work thins. Not because the work is going to disappear tomorrow, but because the people who've already done that thinking will be steadier when the ground moves, and the people who haven't will struggle.

The confession, again

I want to come back to Musk's confession, because I think it deserves more than a dismissive reading.

He said "these are not things I wish would happen." He said "I don't think we're going to have a choice" about universal basic income. He asked out loud how people will have meaning when their labour isn't required.

You don't have to agree with Musk on much to notice that what he's describing is already true in pockets of the economy, and increasingly true in pockets of medicine. The person at the console is still there. Fewer of them, doing more supervision, with more of the task handled by the machine. That's the direction of travel.

The question isn't whether perfusion survives. The profession survives. The question is what the person in the chair is for, and whether they've built a life that can hold the answer when the answer changes.

I don't have a clean ending for you. I don't think there is one. But I know the practitioners who will do best are the ones who can look at this clearly and not flinch. Not because they're optimistic. Because they're prepared.

The machine didn't wait. We shouldn't either.