- Yes, AI is genuinely cutting jobs — about 205,000 of them in the US through August 2026, and it is now the single most-cited reason companies give.
- But the losses are concentrated, not general. Four kinds of work account for most of it: customer service, data operations, entry-level software and finance back-office.
- If your work involves judgement, physical presence, accountability or relationships, you are not in the current wave. That is most jobs.
- The real risk is the training ladder, not the job. Agents absorb exactly the routine work juniors used to learn on — so the danger is fewer entry points, not mass replacement.
- India is not following the US script. TCS alone targeted around 42,000 fresher hires in FY26. The jobs did not vanish; the job description changed.
This question deserves a straight answer rather than either of the two comfortable ones.
The comfortable pessimistic answer is that AI is coming for everything and you should be afraid. The comfortable optimistic answer is that technology always creates more jobs than it destroys, so relax. Both are ways of avoiding the actual data.
The data exists. It is specific. And it says something more useful — and more uncomfortable — than either slogan.
The short answer
Probably not your job — but possibly your first job.
The 2026 layoff data shows AI displacing a narrow band of highly routine, text-bound, well-specified work. If that describes your day, the risk is real and near. If it does not, the near-term risk to your role is low.
The broader risk applies to almost everyone and is subtler: the tasks AI does best are the tasks people used to do while learning. Remove those and you have not removed a job — you have removed a rung.
What the 2026 numbers actually say
Here is the evidence, with sources, because this topic attracts more assertion than measurement.
| What | Figure | Period |
|---|---|---|
| US job cuts naming AI as the reason | ~205,000 | Through August 2026 |
| Same figure at the halfway point | 101,743 | Through June 2026 |
| Whole of the previous year | 54,836 | All of 2025 |
| Share of all 2026 layoffs citing AI | ~22% | Through May 2026 |
| Technology sector cuts | 149,023 (31% of all cuts) | 2026 to July |
Three readings of that table, in order of how much they matter.
First: the trend is real and steep. The 2026 figure passed the whole of 2025 before the year was half over. AI has been the single most-cited reason for job cuts for several consecutive months. This is not a rounding error or a media panic — companies are telling a layoff-tracking firm, in writing, that AI is why.
Second: it is concentrated, not general. Technology alone accounts for around 31% of all cuts. Analysts describe AI's employment impact as "very concentrated" within specific sectors. A national total sounds like weather; it is closer to a localised storm.
Third — and read this one carefully: "cited" is not the same as "caused." These figures come from what employers state as the reason. AI is currently a socially acceptable explanation for a cut that might also be about over-hiring in 2021, interest rates, or a bad quarter. Some of that 205,000 is genuine substitution. Some is a story told about an ordinary redundancy. Nobody, including me, can tell you the split.
That caveat cuts both ways, so hold it loosely rather than using it to dismiss the number.
Who is genuinely exposed
The cuts cluster in work with four traits in combination. Not one of them — all of them.
Work is most exposed when it is:
1. High volume — the same shape of task, many times a day.
2. Well specified — there is a correct answer and a known way to get it.
3. Text-bound — it happens entirely in documents, tickets and screens.
4. Checkable by machine — you can tell it went wrong without a human reading it.
Score all four and the economics of automation are overwhelming. Score two and they are not.
Which is why the named categories are what they are:
- Customer service representatives. High volume, scripted, text, and quality is measurable. The most exposed category in the data.
- Data entry and data operations. The purest case — this is what the four traits describe.
- Entry-level software. Not software engineering. The entry-level slice: boilerplate, small well-specified tickets, test scaffolding.
- Finance back-office. Reconciliation, routine processing, standardised reporting.
If you recognise your day in that list, the honest advice is to act now while you have a salary and time, rather than after a decision is made for you. The section on what to do is written for you specifically.
Who is not in this wave — which is most people
The four traits also tell you where automation stalls. Work resists the current wave when it involves:
- Accountability. Someone must be answerable — legally, professionally, or to a customer. A model cannot hold a licence, sign an audit, or be sued.
- Judgement under ambiguity. No correct answer exists; the job is deciding. Most senior work is this.
- Physical presence. Trades, healthcare delivery, logistics, hospitality, field work. Astra can operate a computer. It cannot rewire a house.
- Relationships and trust. Sales built on a decade of history, negotiation, care work, teaching a nervous person something hard.
- Owning the outcome. The person who decides what should be built, and carries it when it fails.
Notice that none of these are safe because AI cannot help with them. AI helps with all of them. They are safer because the help does not remove the person — it changes what the person spends the day doing.
The real risk: the ladder, not the job
This is the part most coverage misses, and it is the reason I would not tell a 22-year-old to relax.
The work AI absorbs most easily is the work people used to do while learning.
Consider how professions actually train people. A junior developer fixes small tickets for two years and absorbs the codebase. A junior analyst builds the same report forty times and learns what the numbers mean. A junior lawyer reviews documents until patterns appear. A junior marketer writes the ad variants nobody senior wants to write.
None of that routine work exists to produce value. It exists to produce a senior person.
Automating junior work looks like pure efficiency on a spreadsheet. The output still appears, faster and cheaper, and the headcount line improves.
The cost is invisible for about five years — and then it is a missing generation of people who never got the reps. You do not notice a broken training ladder until you need someone at the top of it.
Reporting on 2026 says this directly: junior roles may erode first, precisely because agents absorb the routine learning work that used to train new talent.
So the sharper version of your question is not "will AI take my job". It is:
"If I am early in my career, how do I get the reps when the reps have been automated?"
I do not think anyone has fully solved that. The partial answer is to seek the work that was never on the ladder — problems nobody has specified yet, where the difficulty is deciding what to do rather than doing it.
India and the UAE: a different picture
Almost every article on this topic is written about the American labour market. If you are reading this from India, the numbers you actually need look different — and considerably less alarming than the headlines.
| Signal | What it shows |
|---|---|
| Fresher hiring, major IT firms | ~82,000 graduates expected across TCS, Infosys, HCLTech, Wipro and peers in FY2026 |
| TCS alone | ~42,000 entry-level hires targeted in FY26 |
| Infosys | 20,000+ freshers planned for 2026 |
| BPO / ITES hiring | Up 21% year on year in January 2026, with fresher hiring up 39% |
| The catch | Volume recruitment reducing in H2 2026; shift to "skills-first" hiring and AI-enabled micro-teams |
| The real change | 60–70% of fresher job descriptions now require AI/ML proficiency — including non-AI roles |
Read those rows together, because separately they mislead.
The jobs did not disappear. The job description changed. Indian IT is still hiring graduates in tens of thousands — that is the opposite of the American story. But it is hiring them into smaller, AI-enabled teams, on a skills-first basis, and expecting AI fluency in roles that have nothing to do with AI.
Which means, if you are entering the Indian tech market:
- The door is open. Do not let US layoff headlines convince you otherwise — they are describing a different labour market.
- The bar moved, not the gate. AI fluency is now table stakes for a general software or analytics role, not a specialisation.
- Volume hiring is softening. The trend through the second half of 2026 is fewer, better-qualified hires. Being generically employable is worth less than it was.
- Pay still rewards specialisation sharply. Standard engineering tracks start around ₹3.5–4.2 LPA; specialist and digital tracks run roughly ₹6.25–9.5 LPA. That gap is the return on being specific.
For the UAE the public data is thinner, but the pattern in hiring briefs is similar: demand concentrated in people who can direct and secure AI systems rather than in people doing the work AI now does.
What to actually do about it
Advice that is specific enough to act on this week, ordered by how much it matters.
Move toward the exception, not away from the field. Every high-volume process has a 10% of cases that are messy, escalated, or need a person. That is where the remaining roles concentrate — and they pay better than the routine work did.
Become the person who runs the automation. Somebody has to specify, monitor, correct and be accountable for the system doing the old job. That person is usually recruited from the team that did it manually — but only if they learned the tool early.
For everyone, regardless of role:
- Learn to direct the tools, not just use them. The durable skill is not prompting. It is knowing what good output looks like, catching when it is subtly wrong, and being accountable for shipping it. That requires domain judgement AI does not have.
- Get closer to the decision. Work that ends in "so we should do X" is far safer than work that ends in a deliverable someone else interprets.
- Build things people can verify. A portfolio, a measurable result, a system you ran. As output gets cheap, evidence that you produced a result gets valuable.
- Do not compete on volume. Producing more words, more code, more variants faster is competing directly with the thing that is better at it. Compete on being right.
If you are early in your career, and the routine work you would have learned on is gone: apprentice yourself to a problem rather than a task. Find something nobody has specified. Ambiguity is now the training ground, because clarity has been automated.
The framework I use for deciding what to automate and what to keep human is in automate the process, not the advantage — written about a games studio, but the logic transfers directly to a career.
FAQ
Is AI actually causing job losses in 2026?
Yes, measurably. AI was named in roughly 205,000 US job cuts through August 2026, against 54,836 for the whole of 2025, and it has been the single most-cited reason for cuts for several consecutive months. The important caveat is that these are employer-stated reasons — some of that total is genuine substitution and some is a convenient explanation for an ordinary redundancy.
Which jobs are most at risk from AI?
Work that is simultaneously high-volume, well-specified, text-bound and machine-checkable. In practice that means customer service representatives, data entry and data operations, entry-level software roles, and finance back-office processing. Jobs need all four traits to be seriously exposed — having one or two is not enough for the economics of automation to work.
Which jobs are safe from AI?
Work involving accountability someone must legally or professionally hold, judgement where no correct answer exists, physical presence, relationships built over time, or ownership of an outcome. That covers most jobs. They are not safe because AI cannot help with them — it can — but because the help changes the work rather than removing the worker.
Will AI take entry-level jobs first?
That is the clearest structural risk in the 2026 data. Agents absorb exactly the routine work that used to train new people — small tickets, repeated reports, document review. Automating it looks like efficiency on a spreadsheet, but that work existed to produce senior people rather than value, so the cost appears years later as a missing generation.
Is AI taking jobs in India too?
Not in the same pattern. India's major IT firms expected to onboard around 82,000 graduates in FY2026, with TCS alone targeting roughly 42,000 entry-level hires and Infosys 20,000-plus, and BPO/ITES fresher hiring rose 39% year on year in January 2026. What changed is the job description rather than the job count: 60 to 70% of fresher roles now require AI/ML proficiency, including roles that are not AI-focused, and volume recruitment is softening through the second half of 2026 in favour of skills-first hiring.
What should I do if my job is at risk from AI?
Move toward the exceptions rather than out of the field — every automated process keeps a difficult minority of cases that need a person, and those roles pay better than the routine work did. Then become the person who specifies, monitors and is accountable for the automation, which is usually recruited from the team that previously did the work manually, but only among those who learned the tool early.
Can AI do my whole job or just parts of it?
For almost all roles, parts. Even GPT-6 Astra, the model built to operate a computer directly, reports 72.6% on the OSWorld 2.0 computer-use benchmark — meaning roughly one task in four still fails. That level of reliability replaces tasks and reshapes roles; it does not run a job unsupervised.
- Challenger, Gray & Christmas, monthly job-cut reports, 2026 — the AI-attributed layoff figures
- CNBC, AI is now the leading reason companies give for cutting jobs
- Naukri JobSpeak, white-collar hiring trends, 2026 — Indian hiring and BPO/ITES figures
- The Hire Hub, IT fresher hiring trends India 2026 — fresher intake numbers and the AI-skills requirement
- Jayant Solanki, what GPT-6 Astra actually does — the capability driving this wave
Jayant Solanki
AI-Ready SEO, GEO & AIO strategist based in Indore, India. I work with teams in India, the UAE and the US, which is why this piece refuses to treat the American labour market as the whole story.
More about Jayant →