The book Don't Replace Me argues that AI takes over tasks much faster than it takes over jobs, and unevenly; most people will be fine, and some will not. Dmitry Kargaev wrote this page from the argument in his book Don't Replace Me, and dontreplace.me is his site. A 15-minute check for your own job follows.
Tasks go before jobs
The book's introduction states the idea in one line: "AI takes over tasks much faster than it takes over jobs, and it does it unevenly, company by company, desk by desk." Three outside sources, each a different kind of evidence, point the same way.
One is an employer survey. The US Census Bureau's experimental Business Trends and Outlook Survey ran in 2023 and 2024, and its results were published in Economics Letters in November 2024. Among firms using AI, about 27% reported replacing worker tasks, while only about 5% reported a change in employment because of AI. The authors write that these rates are expected to rise to nearly 35% and 12% in the near future. A slightly higher share of firms reported an employment increase than a decrease. AI use itself began from a low base, 3.7% in September 2023.
Another is a UK Parliament research briefing, POSTnote 757, published on 23 December 2025. It says jobs are more likely to be partly automated than entirely replaced, and that evidence of widespread job loss due to AI is limited, although specific losses have occurred for some roles.
A third is an interview, which makes it opinion rather than data. A Wharton professor told CNBC Make It on 25 July 2025 that AI "doesn't substitute well for human jobs, overall", and advised picking jobs with as many "bundled" tasks as possible.
Where it shows first: young workers in exposed jobs
Young workers in exposed jobs are where the book says the change shows first. The Stanford paper "Canaries in the Coal Mine?" by Brynjolfsson, Chandar and Chen found, in its August 2025 version, a relative drop of about 13% for workers aged 22 to 25 in exposed jobs. The August 2026 revision, with data to June 2026, puts the figure at about 19%. The authors do not claim that AI caused the drop. The UK briefing adds that the early career studies it reviewed also suggest AI is not necessarily the sole cause of declines in early career employment.
Tech Times UK reported on 7 October 2026 that the August 2026 update, built on payroll data for millions of US workers through June 2026, found no sign of job loss across the whole economy. The gap, it says, comes mostly from companies hiring fewer young people rather than firing them. Experienced workers in the same jobs show no gap like this.
Layoff announcements add a second set of numbers. In October 2026, Challenger, Gray & Christmas, which counts announced US job cuts, reported AI cited for 121,401 cuts, about a fifth of all. Total cuts were down 44% on the same months of 2025, and by fall companies were citing the economy more often than AI. The book's verdict is that some of the AI-cited cuts are "the same old cuts with a fresher excuse".
What Goldman's 300 million "exposed" figure means
In March 2023 Goldman Sachs put 300 million full-time jobs in the "exposed" category. The book's point in chapter 1 is that a headline about exposure is not a headline about job loss. The UK briefing describes exposure metrics as estimates of how likely a job is to be affected, based on the tasks AI can perform.
Goldman's US split, as the book reports it, shows how wide the category is: about 7% in jobs where AI could do half the tasks, about 63% in jobs partly taken over, and about 30% in jobs barely touched. The Census paper's summary adds that the Goldman report also said there could be large scale displacement of workers.
This came from a book.
Don't Replace Me
200+ pages. 24 chapters. The honest version of what AI means for your career, written by someone who actually builds this stuff.
Get the Book →Why most people will be fine and some will not
Chapter 3 of the book uses US switchboard operators as its example of people who lose out. In 1950 about 342,000 people worked US switchboards, in the Bell System and independent companies. The US Bureau of Labor Statistics counts about 40,000 operators in 1984 and about 4,600 in 2020.
The book cites Feigenbaum and Gross (2024), who found that operator jobs for young women dropped 50 to 80 percent after the cutover. Operators already in the job who were displaced did worse a decade later. The book keeps the line that "historically, on net, after painful transitions, technology has created more work than it destroyed", and adds that people already in the affected jobs often do not recover.
Step 1: find your appendix
The book's first rule is to find your job's appendix: the tasks, such as status reports, meeting notes and copy-paste work, that you should fear in place of the job itself. The aim is to name them before someone else does.
Start on paper, as the introduction asks. Write your job title and the parts of your work you fear a machine will take. That takes 2 minutes, and the book returns to the paper later.
Then run what chapter 1 calls the distance test, which takes 20 minutes. Find out which AI tools your company pays for or allows, whether a written policy exists, and who uses AI daily. The chapter includes a script for asking your manager or IT. When you are done, add one appendix item to the paper.
Step 2: check which pile you're in
Chapter 3 of Don't Replace Me gives a pile test that takes 15 minutes and has three parts: list your tasks, answer six questions, then count the yeses.
List the five tasks that took most of your hours
Write down the five tasks that ate the most hours in your work, and keep the list in front of you for the next part.
Answer six yes/no questions
Answer each question yes or no.
- Could a chatbot draft it?
- Could a computer check it?
- Could a stranger do it in a month?
- Are you in the first three years of your career?
- Does nobody outside your team ask for you by name?
- Does nobody ask your opinion?
Count your yeses
| Yes answers | Pile |
|---|---|
| 0 to 2 | Bundle pile |
| 3 to 4 | Changing pile |
| 5 to 6 | Exposed pile |
Write the result on the paper from step 1. Chapter 3 adds that jobs made of one checkable task are in real trouble, and that entry-level work is where the change shows now.
Step 3: find your square
Chapter 19 separates two kinds of change. Task automation is what most people face, and the advice there is to keep going. Industry disruption is when AI makes your employer's product free, and the advice there is to move sideways while you are still employed.
The exercise, called find your square, takes 15 minutes. Score your tasks from 1 to 4 and your industry from 1 to 4, then plot the two scores on a 2x2 grid whose squares are named chill, pivot, level up and run. Photograph the grid, date it, and set a reminder to retake it in six months.
What is not known
Nobody can say exactly how many layoffs AI causes. Tech Times UK notes that Challenger counts announcements, so the reason recorded is whatever the company says it is. It adds that some experts call it "AI washing" when a company blames AI for cuts it planned anyway.
The UK briefing says the overall level of AI use in the UK remains uncertain. It also says it is unclear whether AI is creating broad economic gains.
The book states its own limit. Its epilogue says some people who do everything right will still be laid off. What it promises is that you will see it earlier, know your pile and square, have real signed work to show, and be the person who knows how this works.
The longer version: Don't Replace Me
Don't Replace Me: A Survival Guide to the AI Apocalypse, second edition, 2026, is the longer version of this page. It has 22 chapters, each ending in one rule and one exercise you can do the same week. As listed on Amazon on 11 November 2026, the Kindle ebook is $2.99, the paperback (174 pages) is $14.99 and the hardcover is $24.99. The book sells nothing after it: no course, no community, no paid next step. The introduction is a free download at dontreplace.me.
