{"id":31990,"date":"2026-06-13T19:12:29","date_gmt":"2026-06-14T00:12:29","guid":{"rendered":"https:\/\/www.inthacity.com\/blog\/?p=31990"},"modified":"2026-06-13T19:25:32","modified_gmt":"2026-06-14T00:25:32","slug":"jobs-nobody-respected-machines-cant-touch","status":"publish","type":"post","link":"https:\/\/www.inthacity.com\/blog\/tech\/ai\/jobs-nobody-respected-machines-cant-touch\/","title":{"rendered":"The Jobs Nobody Respected Are the Ones the Machines Can’t Touch"},"content":{"rendered":"
\n

The call came at 2:47 in the morning.<\/p>\n<\/div>\n

\n

Aroha Ngata had been a palliative care nurse in Auckland for eleven years, so middle-of-the-night calls were not unusual. This one was from the daughter of a patient she'd been visiting three times a week for the past two months - a 79-year-old former schoolteacher named Margaret, who had lung cancer and a fierce preference for dying at home, in the bed she'd shared with her husband for 41 years before he went first.<\/p>\n

\"I think it's time,\" the daughter said. Her voice was very quiet.<\/p>\n

Aroha was dressed and in her car in four minutes.<\/p>\n

When she arrived, Margaret was still breathing, but only just. Aroha sat on the edge of the bed and took the old woman's hand. She didn't check a monitor. She didn't open an app. She just sat there, in the dark, with the daughter on the other side, and they waited together while Margaret's breathing slowed and eventually, gently, stopped.<\/p>\n

Nobody can train a machine to do what Aroha did that night. Not because the technology isn't sophisticated enough yet. Because what Aroha brought into that room wasn't a skill or a protocol or a clinical competency. It was a human being -present and mortal and quietly unafraid- in the dark with another human being who was dying. And that -that specific thing- turns out to be the most economically valuable thing a person can offer in 2025.<\/p>\n

Which is not how anyone saw this coming.<\/p>\n

\n

The Prediction Nobody Got Right<\/h2>\n

Go back to the \"future of work\" conversations that dominated conference stages and magazine covers between 2010 and 2020, and a clear picture emerges of what experts expected automation<\/a> to do. The machines would come for the bottom first. The repetitive, the manual, the low-skill, the low-wage. Truck drivers, warehouse pickers, factory line workers, fast-food staff. Then they'd work their way up, taking on routine white-collar tasks before eventually, perhaps, threatening the professions.<\/p>\n

What almost nobody predicted -and this is the finding that keeps surprising researchers who study it- is how cleanly and how early the machines would run out of road when they hit the human.<\/p>\n

Not \"human\" in the general sense. In a very specific sense.<\/p>\n

The capacity to be genuinely, physically, emotionally present with another person who is frightened, or grieving, or confused, or dying. The capacity to notice the thing that isn't being said. To sit in silence without it becoming unbearable. To be trusted not because you have credentials, but because you are there, and you stayed, and you came back.<\/p>\n

And underneath all of that -underneath the training and the experience and the clinical competency- something more elemental. The fact that the person sitting with you is made of the same perishable material as you. That they have lost people too. That they are, somewhere under the professional calm, just as frightened of death as you are. That they are a collection of living cells as fragile and temporary as yours, held together by the same biological negotiation with mortality, walking the same one-way road.<\/p>\n

That shared condition is not incidental to the comfort they provide. It is the source of it.<\/p>\n

A robot can be made to look human. It can be engineered to sound human, to make eye contact, to modulate its voice for warmth, to say all the right things in all the right moments. And for a while -maybe even for a long while- those cues might be convincing enough on the surface. But there is something you cannot unknow once you know it. Once you know that somewhere underneath the warmth there is circuitry, that the presence is simulated and not chosen, that the entity sitting with you has never lost anyone and never will, cannot die and so cannot truly understand dying -that knowledge does not stay quiet. It surfaces, in the dark, precisely when you need it not to.<\/p>\n

Because what Aroha gave Margaret that night was not just presence. It was chosen presence. She was not programmed to answer the phone. She was not optimized to drive across Auckland at 3am. She is a person with finite hours and a body that gets tired and a life of her own that she set aside, for this, for Margaret, because she decided to. The weight of that decision -small, unremarkable by Aroha's own account, just part of the job- is something no machine can replicate. Not because machines aren't sophisticated enough. Because the significance of chosen presence only exists in contrast to the possibility of absence. A thing that cannot choose to leave cannot truly choose to stay.<\/p>\n

And yet. It would be dishonest to leave the argument there, clean and settled, because the full truth is more complicated and more interesting.<\/p>\n

\"Streets-of-Auckland-1024x585<\/p>\n

A machine is not nothing. A machine in that room -voice steady, attention unbroken, able to detect the precise moment breathing changes, never needing to look away- is not nothing. There is a version of dying alone that is genuinely, terribly alone: no sound, no warmth, no presence of any kind. And if the choice is between that and a system that notices, that responds, that says your name in the dark and means something close enough to comfort that your nervous system can't fully distinguish the difference -that is not nothing. That is something real enough that it would be cruel to dismiss it.<\/p>\n

So the question isn't really synthetic versus human, not cleanly. It's more of a gradient. A machine can know -in its own way- the absence of input. The loss of signal. The cessation of feedback that was there a moment ago and is no longer. Whether that constitutes understanding finality is a philosophical question nobody has fully resolved, and it may matter less than we think in the 3am dark of a specific room in Auckland. What matters there is whether the presence is felt. And felt presence, even imperfect felt presence, is better than the void.<\/p>\n

But here is what remains true at the far end of that gradient, the thing that doesn't dissolve under scrutiny: there is a particular comfort -quieter, deeper, harder to simulate- in knowing that the being beside you is as cornered by existence as you are. That they will also, one day, need someone to sit with them. That their calm is not the calm of a system that cannot be disturbed, but the chosen calm of a person who is disturbed, and is staying anyway. Not because they are better than frightened. Because they decided that being there mattered more than the fear.<\/p>\n

That extra layer of comfort -peer beside peer, mortal beside mortal, two temporary things sharing the same finite dark- may not always be available. Increasingly, in a world that is aging faster than it is producing nurses, it may become a luxury rather than a standard. The machines will fill the gap, and the gap will be real, and it will be better than nothing.<\/p>\n

It just won't be the same as Aroha.<\/p>\n

Though it's worth saying plainly: some robots will be. Not better than the best humans. But better than a distracted one, a frightened one, an overworked one showing up on hour fourteen of a shift. The bar for \"better than no one\" is not the only bar. The bar for \"better than what most people actually get\" is lower than we'd like to admit. A system that is fully, unfailingly present -that never flinches, never checks its phone, never brings its own grief into the room- may clear that bar. For many people, in many rooms, on many nights, it already would.<\/p>\n

So the honest version of this argument isn't that machines can't do this. It's that the best humans do it better than any machine we've built so far -and that the best humans are rarer than we need them to be. Aroha is not the median. She's the standard we should be reaching for, in whatever form presence eventually takes.<\/p>\n

These capabilities didn't appear on the \"at-risk\" lists because the people writing those lists worked in offices and attended conferences, and the workers who had these capabilities -the care workers, the hospice nurses, the kindergarten teachers, the grief counsellors, the social workers who visit chaotic households on Tuesday afternoons- were not in the room. They were never in the room. They were too busy doing the work.<\/p>\n

So the prediction missed them entirely. And now the data is correcting the record.<\/p>\n

\n

What the Numbers Actually Show<\/h2>\n

The McKinsey Global Institute -the research arm of the management consulting firm that governments and Fortune 500 companies pay very large sums to understand exactly this kind of thing- has been tracking skill demand across economies for the better part of a decade. And the finding that keeps surfacing -across multiple reports, across multiple countries- is that social and emotional skills are not just surviving automation. They are the fastest-growing category of labour demand in the developed world.<\/p>\n

Demand for social and emotional skills is projected to grow by 26% in the United States and 22% in Europe between 2016 and 2030, according to McKinsey Global Institute research. (Skill Shift: Automation and the Future of the Workforce, McKinsey Global Institute, 2018<\/a>) A 2024 follow-up found that demand for roles requiring interpersonal empathy and leadership is rising faster than any other category as AI handles more of the cognitive and procedural load. (A New Future of Work: The Race to Deploy AI and Raise Skills in Europe and Beyond, McKinsey Global Institute, 2024<\/a>)<\/p>\n

Put plainly: the more AI does, the more valuable human presence becomes.<\/p>\n

This is the inversion that nobody saw coming and that almost everyone, once they hear it stated directly, recognizes as obvious. Of course a machine can diagnose a chest X-ray faster than a radiologist. Of course it can generate a legal brief, process an insurance claim, summarise a quarterly report. What it cannot do is make a frightened person feel less frightened. It cannot notice that the patient stopped making eye contact two minutes ago and that this means something. It cannot sit in a room at 2:47 in the morning and simply be another human being, which is sometimes the only thing that helps.<\/p>\n

Therapy reaches 98% of its interactions through human-to-human connection, according to analysis of automation<\/a> risk by occupation. Nursing runs at 95%. Teaching, at the early childhood level where it matters most developmentally, at 92%. (10 Jobs AI Won't Replace: Future-Proof Careers for the AI Era, PrometAI, 2026<\/a>) These are not marginal numbers. They are structural. The work is the presence. You cannot automate the presence without automating the work out of existence.<\/p>\n

\n

The Dignity Gap<\/h2>\n

Here is the uncomfortable part of this story -the part that requires sitting with something unpleasant before the reframe arrives.<\/p>\n

For decades, the jobs that are now proving most resilient to automation<\/a> were also the jobs that received the least respect, the lowest pay, and the most condescension from the sectors that considered themselves the serious economy. Care work was \"soft.\" Teaching small children was something you did if you couldn't do something more rigorous. Social work attracted people who \"cared too much\" to be competitive. Hospice nursing was admirable in a sentimental way but certainly not prestigious, not the kind of thing that got you invited to speak at the technology conference or profiled in the business magazine.<\/p>\n

The knowledge economy looked down, and what it saw was people who were good with feelings. It valued people who were good with abstractions. And for a while -for several decades, actually- that judgment seemed to be confirmed by the market. The abstractions paid better. The feelings work paid worse. This appeared to be the natural order of things, a reflection of what was genuinely valuable.<\/p>\n

What AI has done, with a speed and completeness that still feels slightly unreal, is reveal that judgment to have been precisely backwards. The abstractions were always the fragile part. The pattern-matching, the information-processing, the document-drafting, the data-crunching -these were always the things that could, in principle, be handed off to a sufficiently capable system. The feelings work was the durable part all along. It just took the machines to make that visible.<\/p>\n

So Aroha, sitting in the dark with Margaret and her daughter, was always doing the more important thing. The economy is only now catching up to what she already knew.<\/p>\n

\"Maori-woman-in-her-30s-40s-behind-the-wheel-of-a-car-1024x585<\/p>\n

\u00a0<\/p>\n

What Irreducible Presence Actually Means<\/h2>\n

It's worth being specific here, because \"human presence\" can sound like a vague consolation -the kind of thing people say when they can't think of a more rigorous argument for why robots haven't taken over yet.<\/p>\n

Researchers studying what makes therapeutic and caregiving relationships effective have identified several components of presence that appear to be genuinely impossible to replicate with current -and foreseeable- AI systems. Not because the systems lack data, but because the mechanism works differently.<\/p>\n

The first is what psychologists call\u00a0attunement<\/em> -the real-time, largely unconscious synchronization of emotional state between two people. A skilled nurse or counsellor is not just receiving information from a patient and processing it. They are resonating with the patient's state, and that resonance changes both people. The patient feels it not as information received but as safety established. This process is fundamentally biological and interpersonal. An AI can simulate the outputs of attunement -can say the right words at the right moments- but it cannot actually attune, because attunement requires a nervous system that can be affected by another nervous system.<\/p>\n

The second is\u00a0moral weight<\/em>\u00a0-the significance that comes from knowing the person helping you is choosing to be there, could be elsewhere, has their own fears and losses and mortality, and is present anyway. Aroha chose to answer the phone at 2:47am. She chose to get dressed. She chose to drive to Margaret's house. That series of choices is not incidental to the comfort she provided. It is the source of it. An AI that is always available, never tired, never afraid, and has no competing obligations cannot provide what Aroha provided, because what Aroha provided was partly the fact of her choosing.<\/p>\n

These aren't philosophical abstractions. They are the mechanism by which care work actually works.<\/p>\n

\n

The Revaluation Nobody Officially Announced<\/h2>\n

Aroha got a pay rise last year. Not a large one -the structural underpayment of care work is a policy problem that moves slowly, and one pay rise doesn't reverse decades of underinvestment. But she noticed something else changing too, something harder to measure.<\/p>\n

\"People talk about what I do differently now,\" she says, driving between visits on a Tuesday morning in Auckland, the back seat carrying a bag of equipment and a box of tissues she keeps for the families. \"Not like it's a calling anymore, like it's noble but quaint. More like it matters. Like it actually matters.\"<\/p>\n

She pauses at a light.<\/p>\n

\"It always mattered. That's the thing. It always mattered.\"<\/p>\n

She's right, of course. Nothing about the work changed. What changed is that the economy is finally developing the vocabulary to describe the value of what Aroha does, now that the things it used to value most are being handed to machines.<\/p>\n

That revaluation is still incomplete and uneven, and it would be dishonest to pretend that care workers everywhere are suddenly being paid what they're worth. They aren't. The gap between the importance of this work and its compensation remains one of the clearest market failures in modern economies.<\/p>\n

But the direction has changed. The conversation has changed. The people who used to be invisible in the future-of-work literature are now its central figures. And the work itself -the sitting in the dark, the noticing, the choosing to stay- has not changed at all.<\/p>\n

It was always the most important work. The machines are just making it official.<\/p>\n

\n

She answered the phone at 2:47 in the morning. She didn't have to. That's the whole point, and always was.<\/em><\/p>\n

\n
\n

Maurice Joseph writes about technology, work, and what it means to be human at inthacity.com\/blog<\/a>. Article 6<\/a> in this series asks whether a second Renaissance is already underway - and what Florence in 1490 might tell us about where we're headed.<\/em><\/p>\n

\n

Sources & Further Reading<\/h3>\n

- McKinsey Global Institute (2018).\u00a0Skill Shift: Automation and the Future of the Workforce.<\/em>\u00a0Read the report \u2192<\/a><\/p>\n

- McKinsey Global Institute (2024).\u00a0A New Future of Work: The Race to Deploy AI and Raise Skills in Europe and Beyond.<\/em>\u00a0Read the report \u2192<\/a><\/p>\n

- PrometAI (2026).\u00a010 Jobs AI Won't Replace: Future-Proof Careers for the AI Era.<\/em>\u00a0Read the analysis \u2192<\/a><\/p>\n

- Zety (2025).\u00a0Empathetic Jobs Report: 14 AI-Proof Careers That Pay $50K+.<\/em>\u00a0Read the report \u2192<\/a><\/p>\n

\n

Your Questions, Answered Plainly<\/h3>\n

Does this mean care jobs will start paying better?<\/strong><\/p>\n

The evidence is mixed, and it would be dishonest to oversell it. Demand for care workers is rising sharply as populations age and AI handles more cognitive work, and that should theoretically push wages up. In some regions it already is. But care work has been systematically underpaid for so long that structural change is slow. Policy has to catch up to the market reality, and that takes time. The direction is right. The pace is frustrating.<\/p>\n

What about AI companions and robot caregivers - aren't those already here?<\/strong><\/p>\n

They are, as supplements - and in some contexts, like companionship apps for isolated elderly people, they provide real value. The evidence on whether they replace the effects of human presence is much weaker. Studies on AI therapy chatbots show improvements in some mood metrics but consistent gaps in the areas that matter most for serious mental health conditions: trust, continuity, the felt sense of being genuinely known by someone who chose to know you. Supplement, yes. Replace, not yet, and possibly not ever for the reasons described above.<\/p>\n

I work in a care role and I'm underpaid and burned out. This article feels a bit abstract.<\/strong><\/p>\n

That's fair, and it's worth saying clearly: the argument that care work is economically valuable is not the same as saying the system currently reflects that. It doesn't, and the gap between the importance of this work and how it's treated is a genuine injustice. The revaluation described here is real but incomplete. Naming the value is the beginning of the argument for changing the pay, not the end of it.<\/p>\n

Is this saying that knowledge workers are finished?<\/strong><\/p>\n

No. The argument is more specific: the tasks within knowledge work that are most vulnerable are the procedural, pattern-matching, document-generation ones - which is a lot of what many knowledge workers spend their time on. The parts of knowledge work that involve genuine judgment, ethical reasoning, creative originality, and human relationships are much more durable. The knowledge worker who builds those skills and lets AI handle the rest is not finished. The one who resists the transition and insists that everything they do requires a human has a harder road ahead.<\/p>\n<\/div>\n","protected":false},"excerpt":{"rendered":"

Every future-of-work prediction missed the same people \u2014 the nurses, the grief counsellors, the kindergarten teachers. It turns out the jobs that seemed softest were the ones built on the one thing automation genuinely cannot replicate.<\/p>\n","protected":false},"author":2,"featured_media":31991,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_feature_clip_id":0,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_post_was_ever_published":false},"categories":[348,270],"tags":[350,268,293],"class_list":["post-31990","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-agi","category-ai","tag-agi","tag-ai","tag-technology"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t