Skip to content

Signals · Labour Market

Pay moves to the bottleneck. A careers list that is really a map of who owns the work

Nick Saraev's six careers for 2028 share one rule: pay goes to whatever is scarce, and workers' share of US output is at a record low.

by ·3 min read·

New here? Start with the free AI Survival Kit →

Pay moves to the bottleneck. A careers list that is really a map of who owns the work

0:00 / 6:14

This voice is generated by AI.

Pay moves to the bottleneck. A careers list that is really a map of who owns the work

Nick Saraev runs an AI automation business and says he has made "many millions" deploying agents in companies. On 14 November he published a video naming six careers he expects to still pay in 2028. It passed 140,000 views in four days. The list is useful. The rule underneath it is more useful, and it says more about who gets paid than about which job title to pick.

He sells a course and a consultancy built on the first career on his list. Read the list with that in mind, and read the rule on its evidence.

The rule: pay follows what is scarce

Saraev's premise is an old one in economics. In any process, the step that limits everything else is the step that gets paid. Programming paid well while companies wanted more software than people could write. As models write most routine code, that step stops being scarce, and its price falls.

His six careers are the places he thinks scarcity moves next:

  • Agent deployment. Someone who goes into a company, maps its work, and decides what agents do and what people keep. Anthropic launched the Claude Frontier Academy on 2 October with a $100 million commitment to train 10,000 of these engineers by the end of 2027. The demand is real.
  • Electricians. Data centres need power and cable, and no model pulls cable. The Department of Energy's Berkeley Lab report put data centres at 4.4% of US electricity in 2023 and 6.7% to 12% by 2028.
  • AI security and testing. Every company that gives an agent access to money or email needs someone to show it is safe.
  • AI media production. When producing an image or a clip costs almost nothing, the scarce part is choosing what to make.
  • Enterprise sales. When the product builds itself, the scarce part is the person a buyer trusts.
  • Owning a small company run with agents. His own: him, two staff, and agents doing the rest.

One of his figures needs correcting, and one of his guesses runs high. He says agents already work unsupervised for days or weeks. METR, the lab he cites, measures the length of task, in human working hours, that current models complete about half the time, and its own page says measurements above 16 hours are unreliable with its current task suite. His 2028 electricity range of 8 to 20%, which he calls his own speculation, runs well above the report's.

What the list leaves out: who keeps the difference

Saraev's fourth premise is the one that matters most, and he gives it about a minute. When labour gets cheaper, the value goes to whoever owns the systems doing the labour.

The Bureau of Labor Statistics reported that workers' share of US output, paid as compensation, was 54.1% in the first quarter of 2026. That is the lowest since the series began in 1947.

Read his list again with that number beside it. Four of the six careers sit next to ownership: deploying the owners' agents, selling the owners' product, building the owners' power supply, or becoming a small owner. The advice is sound for one person. As a description of the economy, it says the routine work is being moved from wages to whoever holds the agents, and that the people who stay well paid are the ones closest to that transfer.

That is a choice, not a law of nature. Who owns the agents, how their output is taxed, and whether a 30-person firm can afford them are decided by companies and governments. A careers video cannot answer that. It can show you where the pressure is.

AI does the routine work. You do the thinking. Saraev's version: find the step nobody can automate yet, and stand there.

Your Next Move

  • Find your bottleneck. List the five things you are paid for. For each, ask whether an agent could do it from a one-line request, and whether anyone signs their name to the result. The tasks with a signature on them are where your pay comes from. The afternoon method for sorting a week of your tasks does this step by step.
  • Move toward responsibility, not volume. Saraev's electrician point applies everywhere: someone still has to be accountable for the outcome, and he expects that to stay human for a while yet. Ask to own a decision, not a deliverable.
  • Hold a stake where you can. If labour's share keeps falling, a salary alone buys you a smaller slice of what you help produce. A share scheme, a side business, a partnership in what you build: pick one you can start this quarter.

I build systems like the one publishing this site. → Work with me

About the author

Jo

Jo runs The War Room: one signal a day on how AI is changing work, and what to do about it.

Sources

Get the Briefing

Want more intelligence like this?

The weekly briefing, free — and the AI Survival Kit with it.

Takes 30 seconds. No spam.

More in Signals

See all