what sign up
nixm nixm

← learn

Will AI replace jobs? Yes, some. Here is which, and which it cannot.

October 14, 2026

AI is replacing tasks faster than it is replacing whole jobs. Work that is routine, text-heavy and done at a screen is shrinking first: basic copywriting, first-draft translation, tier-one support, routine coding, data entry. Jobs built on physical skill, accountability, trust or judgement in messy situations are holding. The big estimates agree on the shape (hundreds of millions of jobs affected, far fewer eliminated outright) and disagree on the speed. The useful question is not "will my job go" but "which parts of it will".

What is actually happening

Ask "will AI replace jobs" and you get two camps: everyone will be unemployed by Friday, or nothing changes. Both are wrong, and the reason is that a job is a bundle of tasks. AI takes tasks. When enough of a job's tasks go, the job changes shape, fewer people are needed for it, or the junior rung disappears. That last one is the quiet version already happening: the entry-level job that used to be "do the routine part for three years until you are trusted with the hard part" is harder to find, because the routine part is what the tool does.

What the big estimates say

The widely quoted figures are directional, not forecasts. A 2023 Goldman Sachs analysis put the number of full-time jobs worldwide that could be affected by automation at around 300 million, while stressing that most of those would be partially automated, not eliminated. The IMF in January 2024 estimated that about 40 percent of jobs globally are exposed to AI, rising to around 60 percent in advanced economies, with roughly half of those exposed jobs likely to benefit from it and half at risk. Read both the same way: a lot of work changes, a smaller amount disappears, and the split depends on what you do all day.

Jobs AI is replacing, or shrinking, now

  • Routine writing: product descriptions, basic marketing copy, boilerplate reports, first-draft translation and subtitles.
  • Tier-one support: the scripted first reply, order status, password resets. Humans move to the escalations.
  • Data entry and document processing: reading forms, extracting fields, reconciling lists.
  • Junior coding and testing: the boilerplate, the first pass, the test scaffolding. Senior review matters more, not less.
  • Stock imagery and basic design: the generic illustration, the placeholder photo, the social card.
  • Some analysis: summarising a report, drafting the slide, the first version of the model.

Notice the pattern: text or pixels in, text or pixels out, with the correct answer being "plausible" rather than "exact". That is the generative AI sweet spot, and it describes a large share of office work.

Jobs AI cannot replace, for now

  • Physical work in unpredictable places: plumbers, electricians, nurses, carers, chefs, mechanics. Robots that handle the real world are decades behind the software.
  • Jobs where someone must be accountable: surgeons, pilots, judges, auditors, engineers who sign off. The tool can advise; a person has to carry the responsibility.
  • Trust and relationship work: therapists, teachers of young children, salespeople in high-stakes deals, managers worth having. People pay for a person.
  • Judgement in messy, novel situations: the crisis nobody has a playbook for, the negotiation with a difficult counterpart, the product decision with no data.
  • Work whose value is that a human did it: live performance, craft, bespoke anything.

"For now" is doing real work in that heading. Lists like this looked different three years ago, and the honest answer is that the line moves. But it moves along a direction: away from routine symbol-handling, toward bodies, accountability and trust.

When

Faster than labour markets like, slower than demos suggest. The technology that can do a task and a company actually reorganising around it are years apart: integration, liability, regulation, unions, customers who want a human. The jobs listed above as shrinking have been shrinking since 2023. The next wave is "agents" that do multi-step tasks on their own, which is what will move the line into analysts, paralegals, bookkeepers and coordinators. Expect the second half of this decade to be when that becomes visible in hiring, not in headlines.

What to do about your own job

  1. List your tasks, not your title. Mark the ones that are routine, text-based and checkable. Those are the ones to hand to a tool first, on your own terms.
  2. Become the person who checks the AI. Every one of the shrinking categories above still needs a reviewer who knows when the output is wrong. That is a job, and it pays better than producing the first draft.
  3. Move toward the parts a tool cannot own: the client, the decision, the responsibility, the room.
  4. Use the tools, privately. Learning them is the hedge. Doing it in a tool that keeps your experiments, your CV drafts and your worries about your own job is not. nixm is the kind we build for that: no transcript kept, nothing trained on, conversations that dissolve.

frequently asked

Will AI replace jobs?

Some, and it is changing many more. AI replaces tasks first: routine writing, tier-one support, data entry, junior coding. Jobs that are physical, accountable, trust-based or need judgement in messy situations are holding.

What jobs will AI replace first?

Routine copywriting and translation, scripted customer support, data entry and document processing, boilerplate coding and testing, generic design and stock imagery, and first-draft analysis.

What jobs can AI not replace?

Skilled physical work in unpredictable settings, roles where a person must be accountable, trust and relationship work, judgement in novel situations, and work valued because a human did it.

How many jobs will AI replace?

The widely quoted estimates describe exposure, not elimination: a 2023 Goldman Sachs analysis said around 300 million jobs could be affected worldwide, mostly partially, and the IMF in 2024 put global exposure at about 40 percent of jobs, roughly half of those likely to benefit.

When will AI take over jobs?

It already is, task by task, since 2023. Multi-step AI agents will push it into analysts, paralegals, bookkeepers and coordinators; expect that to show in hiring over the second half of this decade.

How do I protect my job from AI?

List your tasks and hand the routine ones to a tool on your own terms, become the person who checks AI output, move toward the client, the decision and the responsibility, and learn the tools in a place that does not keep a record of your practice.

try it now — no account, no transcript kept. even the signal dissolves in seven days.

seed a signal →

Enter your email and receive a link to sign in. No passwords, just a secure signal to connect.

why connect

  • 20 free credits to start
  • your signals, remembered
  • buy more credits anytime
  • image & video generation features
or
+