What AI Is Actually Doing to Your Job Market

Goldman Sachs just put a real number on the thing your generation has been whispering about for two years. In a U.S. Daily Note published April 6, 2026, Goldman economist Elsie Peng reported that AI is erasing roughly 16,000 net U.S. jobs a month, pulled straight from payroll data. That’s about 192,000 jobs over the last twelve months. And the workers getting hit hardest are the ones ages 22 to 25.

I want you to read that number and not spiral.

Because the story under the number is more useful than the number itself. AI isn’t evenly erasing jobs. It’s collapsing a specific rung of the career ladder while quietly building new rungs your parents don’t know how to see yet. If you understand which is which, you make very different decisions at 22 than the kid next to you who only read the headline.

Sit down. Let’s walk through it.

The short version

If you only read this far, take this with you.

What the data says	What it means for you
AI is cutting ~16,000 net U.S. jobs per month (Goldman Sachs, April 2026)	The entry-level rung is shrinking, not the whole ladder.
Entry-level postings down 35% since January 2023 (Revelio Labs)	Fewer doors. The ones still open want more proof.
Workers 22-25 in AI-exposed fields: 16% relative employment decline (Stanford)	The hit is concentrated on your exact age bracket.
Entry-level software dev jobs for 22-25 year olds: down ~20% since 2024	Coding bootcamp alone isn’t the shield it was in 2019.
Gen Z excitement about AI dropped 36% → 22% in one year (Gallup/Walton Family Foundation)	Your peers are panicking. That’s an opportunity.

Read that once, then read it again. Three of those numbers describe a shrinking door. One of them describes a stampede of panicked 22-year-olds. That combination is the whole game. When everyone else runs the same direction, the rare thing is to think clearly.

What the Goldman number actually measures

Let me translate the headline before we do anything with it.

Goldman’s 16,000-a-month figure is net. Peng’s team found AI substitution (where a machine replaces human work outright) destroyed roughly 25,000 gross jobs per month over the past year. AI augmentation (where the tech makes an existing worker more productive, letting companies hire more of them) added back about 9,000 per month. Subtract and you get the minus-16,000.

The roles taking the heaviest hits are the ones you’d guess: data entry, customer service, legal support, billing, junior coding. Research from Brynjolfsson, Chandar & Chen at the Stanford Digital Economy Lab adds occupation specifics — operations managers, accountants, auditors, receptionists, information clerks. If the job is mostly routine pattern-matching on a screen, a model can now do a meaningful slice of it.

The roles gaining are harder to see because they’re newer, weirder, and less named. AI-native product roles. Implementation engineers. People who translate between executives and tools. Hands-on trades. Skilled care work. Anything where a human body has to be physically present and accountable for a result. Trade school is one whole conversation in itself, and it’s not an accident that Gen Z started looking at the trades the same quarter the entry-level desk job started evaporating.

That’s the shape of this shift. Not a wall. A rearrangement of doors.

Why 22 to 25 is getting it worst

The honest reason is a little brutal. Your age bracket is concentrated in the exact kind of role AI is best at: the “junior does the routine task, senior handles judgment” model that ran most white-collar training programs for forty years. Pull out the junior rung, the one closest to automation, and the senior rung doesn’t disappear — it just stops needing a new hire to feed it.

Goldman’s payroll data shows entry-level software developer employment among 22-to-25-year-olds is down roughly 20% since 2024. SignalFire’s 2025 State of Tech Talent Report found new grad hiring at Big Tech fell 25% from 2023 to 2024 and kept sliding through 2025. Research firm Revelio Labs found entry-level job postings overall are down 35% since January 2023, with AI a large driver.

Meanwhile — and this is the part nobody tells you — mid-career and senior roles in those same fields have mostly held steady or grown. The market isn’t collapsing. The first-year door is.

That’s a specific problem that needs a specific response. The rest of this post is that response.

What AI actually changed about entry-level work

The World Economic Forum released a January 2026 briefing called “How AI Is Changing Early Careers”, built on interviews with 200+ leaders and a PwC survey of 9,394 entry-level workers across 48 economies. Their finding is less dramatic than the headline and more useful.

AI isn’t deleting entry-level work wholesale. It’s changing what counts as entry-level work. The task that used to get a 22-year-old hired — “research this, summarize it, draft a first pass” — is the exact task AI now handles in 90 seconds. So the bar moved. What companies now want from their junior hires is judgment about the output, taste, follow-through, the ability to push back on the model when it’s wrong, and the social skills to work with humans who pay for the outcome.

The same PwC survey found that 76% of entry-level workers say job security is the single most important factor in a job, but only 53% feel very secure where they are. That gap — wanting stability, not having it — is the emotional temperature of your generation right now. It’s real. Name it honestly. Then don’t let it decide your next move for you.

Here’s the quieter finding in the briefing that I want you to underline: entry-level workers who report using AI regularly are more curious (47%) and more excited (38%) than worried (29%) about the tech’s impact on their work. The kids who lean in aren’t the ones most afraid. The kids most afraid are the ones standing still.

That tells you everything about who’s going to be OK.

What AI didn’t change

AI changed the task. It didn’t change the fundamentals of getting and keeping a job.

A hiring manager in April 2026 still wants to know three things about you — the same three they wanted in 2006, and 1986, and 1966. Can I trust you to show up? Can I trust you to finish the thing? Can I trust you in the room with a client?

I wrote a whole post on this — the thing that actually gets you hired — because the data keeps saying it out loud and your peers keep not hearing it. A 2026 Robert Half survey found that 71% of working professionals rank time management and punctuality as the single most important trait for a new hire. AI skills came in fourth, at 36%. Not because the tools don’t matter. Because everyone has the tools now. Tools are table stakes.

The differentiator is the part AI cannot imitate: a human who consistently does what they said they’d do. That’s more rare in 2026 than it was in 2016. That makes it worth more, not less.

The honest playbook for 2026

Here’s what I’d tell you if you were sitting on the porch next to me asking what to do. Not doom. Not hype. The actual moves.

1. Treat AI like a power tool, not a threat or a religion

Use it every day. For homework. For drafting emails. For unstuck moments at 11 PM. Learn its failure modes — where it lies, where it hallucinates, where it confidently wastes your time. The people who win this decade aren’t the ones who refused the tools or the ones who worshiped them. They’re the ones who used them constantly, noticed the seams, and trusted their own judgment over the output. You can’t get that from a prompt-engineering course. You get it from reps.

2. Get hired on anything while you’re looking for the thing

The ZipRecruiter 2026 data is almost embarrassing in how clean it is: grads with any prior work experience get hired at 81.6%. Grads with none get hired at 40.7%. Lifeguarding counts. Retail counts. A ten-hour-a-week campus gig counts. The gap between “had a job” and “didn’t” is nearly double the outcome. Do not sit at home polishing a resume for six months waiting for the perfect first offer. That story ends badly.

3. Pick a direction that wants a human body or a human judgment

The roles AI is least eating into over the next decade cluster around two things: physical presence (trades, hands-on medical, skilled construction, in-person care) and accountable judgment (anything where someone has to stand up and say “I made this call and I own it”). Both are unfashionable. Both are insulated. The middle — routine desk work with neither physical presence nor owned judgment — is the soft belly.

4. Build a portfolio, not a resume

Your resume says you were at a place. Your portfolio says what you can do. For any field you care about — writing, design, code, marketing, research, sales — have a folder of real things you made, shipped, or sold. Not assignments. Real artifacts. In a world where everyone’s resume was polished by AI, the resume is noise and the proof-of-work is signal.

5. Start money early on purpose

AI isn’t just changing hiring. It’s changing the shape of a career. Fewer clean 40-year ladders, more messy stretches of contract work, gaps, pivots, and layoffs. The only way to make that shape survivable is to stack cash and invest early, so a six-month gap doesn’t turn into a financial emergency. The playbook is how to start investing before you turn 22. Start the paycheck you don’t need yet, so that when you do need it, it’s already there.

6. Don’t take a four-year bet without counting the cost

A 22-year-old graduating in April 2026 with $38k in debt into a labor market where the entry-level door is narrower than it was when they applied is making a specific bet. Sometimes it’s the right bet. Sometimes it isn’t. The college decision post walks through the real math. Read it before you borrow. Re-read it before you borrow more.

What this looks like on a Tuesday

Picture two 23-year-olds, same town, same degree, Class of 2025.

Kid A read the headlines, decided the world was ending, and has spent six months on the couch applying to 600 jobs through portals while refreshing TikTok for AI-panic content. No part-time work. No portfolio. Resume is AI-polished and so is every other resume in the stack.

Kid B took a 15-hour-a-week gig at a coffee shop the month after graduation, started a free Substack writing about his industry, learned three AI tools cold, showed up early to every networking event in town, and sent follow-up emails within 24 hours every time. He got his actual first-choice job eleven months after graduation, which felt like a long time at the time and will look like a short time in ten years.

Neither of them has a better brain than the other. One of them made a decision about who to be while the door was narrow. The other waited for the door to widen on its own.

The door is not going to widen on its own.

The thing I want you to hear

You are walking into a labor market that’s harder to crack than the one I started in. That’s true. I’m not going to pretend it isn’t. And I’m also not going to hand you the “everything will be fine” speech, because you’d smell it from a mile away.

Here’s what I can hand you. The same boring, old-school traits that have always produced a good career still produce a good career — maybe more so now, because fewer of your peers have them. Dependability. Curiosity about the tools. A willingness to do small work well. The guts to pivot when a path closes. Friends who are building something too. A bank account that gives you room to say no.

None of that is replaced by AI. All of it is amplified by AI, in the people willing to use the tools. You don’t need to out-compete a model. You need to be the human the model works for — the one holding the judgment, the relationship, and the outcome. That job is not going anywhere.

The structure of your twenties is the same as mine was, even though the weather looks different. Build the skills. Keep your word. Stack a little money. Pick one direction at a time. Fail cheaply. Move on quickly. Most of the daily operating system that works predates the iPhone, let alone ChatGPT. The tools are new. The life is not.

What to do this week

Five moves. None of them take permission.

1. Audit your last six months honestly. Did you build anything shippable? Did you work a real job, even part-time? Did you learn one new tool cold? If the answer to two of those is no, your fix for this market isn’t another resume polish. It’s starting one of the three.
2. Pick one AI tool and use it every day for 30 days. Not five tools. One. Go deep. Find its limits. Know where it lies. This is your 2026 baseline and employers can tell in five minutes whether you have it.
3. Get any part-time work inside two weeks if you don’t have it. The statistical hit of going from “no work” to “some work” is larger than almost any resume improvement you can make.
4. Start a public piece of work. Substack, Github repo, YouTube, a portfolio site, whatever. One thing that shows taste and follow-through. The entry bar is embarrassingly low. Most of your competition won’t do it.
5. Read the WEF briefing once. The official document is here. Twenty minutes. It will calm you down and sharpen you at the same time, which is rare.

The thing worth remembering

AI is changing the job market. It is not ending the concept of work, or human judgment, or careers that matter. The panic in your feed is louder than the shift in the data, and the shift in the data is louder than the average 22-year-old realizes.

Your job isn’t to beat the machine. Your job is to be the person worth paying to stand next to it. That’s always been the job. It just got more obvious.

Now go do something today that future-you will be glad you started.