Anthropic, the company behind Claude, published its own data on which jobs AI is replacing right now. The office roles everyone was told to chase are in the red zone. The trades everyone was told to avoid are the safe ground. Here is the map, the hiring lens, and what to do with it.
Think about the advice you grew up with. Go to college. Get the office job. Work in finance. Do not become a plumber.
Now look at whose work AI does first.
The company that builds one of the most capable AI systems in the world just told you which jobs it replaces. Not a think tank. Not a newspaper. The source of the technology, publishing its verdict on its own effect. So it is worth reading closely.
What Anthropic Actually Found
Anthropic mapped job categories by how much AI is already doing the work. The report splits roles into zones.
The red zone is active replacement. These are jobs where AI substitutes for human labor at measurable scale today. The blue zone is pressure. Still mostly human, still yours for now, but the share of the work a person does is shrinking.
The part that stings: the red zone is full of the jobs that came with degrees. The ones with career ladders. The ones sold to you as safe.
There is no more credible institution to name which jobs AI replaces than the one selling the AI. And it published the list anyway.
The Two Zones: What Is Red, What Is Blue
Here is the map. Print it, or keep the tab open.
| Zone | Job categories | Status |
|---|---|---|
| Red (active replacement) | Business operations, finance and accounting, software and coding, marketing, some architecture and engineering | AI substitution measurable now |
| Blue (pressure zone) | Paralegal work, data analysis, administrative roles, customer support, content production | Human-led but declining share |
| Low displacement | Trades, physical installation and maintenance, elder care, complex relationship work, owner-operator businesses | Durable for the foreseeable future |
The pattern is consistent. The more a job is pattern recognition, document processing, and knowledge retrieval, the more exposed it is. The more a job needs physical presence, trust, and real-time judgment in messy conditions, the safer it stays.
Read your own week against that line. How much of what you did was moving information around? How much needed you in a room, on a site, or holding a relationship together?
The first number is what AI reprices. The second is what it cannot touch yet.
The Parental Advice Flip
The story of the next decade is not "AI is coming for your job." It is more specific than that, and more uncomfortable.
For forty years the script was one line. Get the degree. Work in finance. Go into law. Write software. Avoid the trades. AI is inverting that script in real time.
The jobs that demanded the most schooling, that carried the most status, that were sold as the safe bet, are the ones where AI delivers the most output for the least cost. A junior financial analyst produces a set amount of work per hour. Claude produces the same in minutes, at a fraction of the cost, with no sick days and no review cycle.
The plumber cannot be replaced by a language model. Neither can the electrician. The founder running a physical-service business is, right now, in a structurally safer position than a mid-level knowledge worker at a white-collar firm.
That is the flip. The prestige jobs are the exposed ones. The hands-on jobs are the durable ones.
The edge is not physical work alone, and it is not AI skill alone. It is the combination. The person who pairs an irreplaceable human element with AI leverage is the one the market rewards. If you want the sharpest version of that combination as a role, the highest-paid new AI job needs zero code, and it is built entirely on this idea.
What This Means If You Are Hiring
If you open roles, run this lens before you write the job description.
Ask two questions about the role:
- What percentage of the work is document processing, knowledge retrieval, or pattern recognition?
- What percentage needs physical presence, trust-based relationship continuity, or context that changes minute to minute?
If the answer to the first is sixty percent or more, that role is being repriced by AI right now, whether you fill it with a human or not.
Here is what to do with that.
Hire operators, not processors. The hire is the person who can point AI at the pattern-recognition work and spend their own hours on judgment and relationships. The person who does the pattern-recognition work by hand is being priced out.
Build the AI layer before you post the role. Do not hire a person and then check whether AI applies. Figure out the AI layer first. Then define what a human adds on top of it.
Value physical and relational skills out loud. In a market flooded with cheap digital labor, a person who can hold a room or fix a thing on site is genuinely scarce. Pay for it like it is scarce.
What to Tell Your Kids
The standard career talk is calibrated to 2005. Telling a teenager to chase a knowledge-worker path because it pays well and it is stable no longer holds for the red-zone roles. The stability assumption is the part that broke.
Better framing for the conversation:
Work where you are the one deploying AI, not the one being replaced by it.
Physical skill plus AI literacy is the rare pairing. An electrician who uses AI to run estimates, write client proposals, and manage the schedule is close to impossible to displace. He owns the part AI cannot do and rents the part it can.
Owner-operator thinking beats credentials. The ability to build and run a business, even a small one, is more durable than a job title in a field AI is eating.
And learn how AI actually works. Not to become a developer. To become someone who can direct AI systems to do useful work. That is the skill the next decade pays for.
Read This Before You Panic
A map of risk is not a prophecy, so hold it at the right distance.
Adoption is slow. The red zone is real, but it is not filling overnight. A finance role in the red zone today is still a finance role next quarter. What changes is the price of doing it the old way, and the speed at which the person who adds AI pulls ahead of the person who does not.
This is also one company's read, drawn from what Anthropic sees inside its own product. Treat it as the best available signal, not as a settled forecast. Your own industry may move faster or slower than the chart suggests.
And the line is not clean. Plenty of red-zone roles keep a human core that AI cannot reach. A great marketer is not just producing copy. A strong engineer is not just writing code. The safe move inside an exposed field is to run toward that human core and let AI take the rest.
The honest version is this. The slow adoption is the gift. It means the gap between the people who understand the shift and the people who do not is still wide, and still yours to cross. In five years it will not be.
If You Do One Thing This Week
Take your own role, or the next role you plan to hire, and split it in two. On one side, the document-processing, retrieval, and pattern work. On the other, the presence, trust, and judgment work.
Then move yourself toward the second column and hand the first to AI. Not all of it. One task. Start with the one you dread most.
That is the whole shift in miniature. The window to build on the right side of this line is open now, and it costs you an afternoon to start.