How to Prepare Your Career for AI Job Automation in 2 Years
A practical guide to AI job disruption: build a personal AI eval, shift toward owning assets, and protect your income before automation hits.
What’s the real risk AI poses to your career right now?
The risk isn’t a sudden mass layoff. It’s a slow, compounding erosion of entry-level opportunity. Research cited from Stanford on AI economics found that employment for 22 to 25 year olds in the most AI-exposed jobs (white-collar roles like writing, reporting, basic accounting, and content production) has dropped 19% below where it would be if it had kept pace with less-exposed jobs. That’s not a future prediction. That’s already happened. Juniors are the first to feel it because the tasks they used to cut their teeth on are the easiest for AI to absorb, while senior employees stick around because they can still judge the quality of AI output.
TL;DR
- Entry-level white-collar work is already shrinking, with AI-exposed jobs for young workers showing measurably lower employment growth than less-exposed roles.
- Building a personal AI eval (a fixed set of real tasks from your job, fed to each new model release) is the only reliable way to track whether AI can actually do your work, since secondhand opinions go stale fast.
- Ownership matters more than income as AI makes labor cheaper: real estate, infrastructure, compute, and businesses hold value in a deflationary labor market in a way that wages won’t.
- Distribution (an audience, network, or reputation) is a portable asset that survives job loss or business failure and helps you pivot faster than starting from zero.
- Physical, hands-on work carries a longer runway than knowledge work, though it’s not permanently automation-proof either.
- Identity built around a job title is fragile, while identity built around character traits and principles holds up better as AI capabilities compress what used to take humans years to learn.
- Basic security prep (food, water, cash, MFA) is cheap insurance against short-term disruptions tied to growing AI-driven cybersecurity risk, separate from the job question but worth handling at the same time.
Everyone else built a construction worker.
We built the contractor.
One file at a time.
UI, API, database, deploy.
How do you know if AI can already do your job?
Stop asking other people and start testing it yourself. Capability claims about any given model age out within weeks, and nobody outside your specific workflow can tell you whether a new release handles your actual tasks. The fix is building what’s effectively a personal benchmark: a standing folder of 10 to 15 representative tasks from your actual job, along with the context needed to complete them (project files, data, templates, whatever you’d normally pull from). Every time a new model ships, run it against that same set and see what sticks.
This matters because the gap closes faster than most people expect. One practitioner tracking this over a three-month stretch saw a model go from completing less than 10% of their task set to over 50%. That’s not a hypothetical curve, that’s a real jump in a single quarter. If you’re not running this test yourself, you’re relying on marketing claims or secondhand Twitter threads, neither of which reflects your specific job.
Why does owning assets matter more than earning a salary?
Automation’s entire economic function is making labor cheaper. That’s true by design, not as a side effect. When labor gets cheaper, the things that don’t get cheaper become relatively more valuable: the infrastructure AI runs on, the compute, the real estate underneath the data centers, the businesses that sit on top of all of it. In a deflationary labor environment, cash flow tied to your personal output becomes a more fragile position than an ownership stake in something durable.
Practically, this means shifting the center of gravity of your financial life away from “what do I get paid” and toward “what do I own.” That can mean real estate, a stake in a business, compute capacity, or any asset that produces value independent of your personal labor hours. Debt financing these positions becomes more attractive in this scenario too, since borrowing costs tend to fall as the broader economy shifts toward automation-driven deflation. None of this is a guarantee, and it carries real risk, but the logic is straightforward: wages compress, assets don’t.
What is “distribution” and why does it protect you?
Distribution means having an audience, a network, or a reputation that follows you independent of any single job or employer. It’s not quite the same as owning a hard asset, but it behaves like one because it’s portable. If you lose a job, distribution gets you in front of the next opportunity faster. If you run a business and it collapses because AI erodes its margins, distribution gives you an existing market to sell the next thing to.
Plans first. Then code.
Remy writes the spec, manages the build, and ships the app.
This is distinct from traditional career advice about “networking.” It’s closer to owning a channel, you have direct reach to people who trust your judgment, and that reach doesn’t evaporate when a company restructures or a market shifts. In an environment where business models can become obsolete in a single product cycle, that kind of durable reach is one of the few things that survives the transition intact.
Is hands-on, physical work safer from automation?
For now, yes, relatively. Work that requires physical presence and manipulation of the real world, plumbing, nursing, skilled trades, carries a longer runway than deskbound knowledge work, simply because robotics and physical automation lag well behind software automation. That doesn’t mean physical work is permanently immune. It means the timeline is longer, likely by a few years rather than a few months.
If there’s a reasonably direct path from your current skill set into physical, hands-on work, it’s worth scouting that path now rather than waiting until the white-collar side of the market tightens further. This isn’t about abandoning a career overnight. It’s about understanding where the insulation currently exists and treating that insulation as temporary, not permanent.
How should you think about identity as AI capabilities keep expanding?
The people who struggle most with rapid AI progress tend to be the ones whose sense of self is fused to a job title: “I’m a lawyer,” “I’m a programmer,” “I’m an analyst.” When systems arrive that outperform humans at the core tasks of that role, that identity takes a direct hit. A more durable approach is defining yourself by traits and principles rather than economic output: whether you’re curious, whether you like solving hard problems, whether you’re someone who takes care of others. Those traits don’t become obsolete when a model ships a new capability.
This matters because the pace of change isn’t slowing down. Anthropic’s CEO Dario Amodei has written that the next five years could compress what would normally be 50 to 100 years of biological progress. Whatever you think of that timeline, the broader point holds: a lot of assumptions people hold about work, status, and lifestyle were formed in a pre-AI world and may not survive contact with a post-AI one. Revisiting those assumptions now, rather than being blindsided by them later, is cheap insurance.
Frequently Asked Questions
What is an “AI eval” and why does it matter for job security?
An AI eval is a personal, standardized set of 10 to 15 real tasks from your job, run against new AI models as they release. It tells you concretely what percentage of your actual work AI can already do, rather than relying on general claims about model capability that may not apply to your specific tasks.
Which jobs are most at risk from AI automation right now?
White-collar, entry-level roles involving writing, reporting, document production, and basic accounting show the clearest early impact, with measurable drops in employment growth for younger workers in those fields compared to less-exposed roles.
Should I quit my job and start a business because of AI?
The advice here isn’t to quit abruptly. It’s to shift focus toward building ownership (real estate, infrastructure, equity in a business) and distribution (audience or reputation) alongside your current income, so you’re not solely dependent on a paycheck if your role gets automated.
Is physical, hands-on work completely safe from AI automation?
- ✕a coding agent
- ✕no-code
- ✕vibe coding
- ✕a faster Cursor
The one that tells the coding agents what to build.
No. It has a longer runway than white-collar knowledge work because robotics and physical automation lag behind software automation, but it’s not permanently immune. Treat it as a few extra years of insulation, not a permanent exemption.
How do I start protecting my career against AI disruption today?
Build a personal task eval and test new models against it regularly, start shifting any spare capital toward owning assets rather than just earning income, cultivate some form of portable distribution like a network or audience, and reassess whether your identity is tied too tightly to a specific job title.
