AI is remaking the student job market. Here's how to keep up.
AI is quietly reshaping who gets hired out of college and who doesn't. A clear-eyed look at what's actually happening, what the fear gets right and wrong, and five things you can do this week to get an edge.
The fear is real. It's also badly framed.
If you're graduating in the next two years, you've probably had the thought: "Is there even going to be an entry-level job for me?"
Your feed is full of it. Layoffs at FAANGs. CEOs quoted saying they "won't hire juniors anymore." 22-year-olds on TikTok crying over 400 applications and no responses. A finance blogger claiming "Goldman doesn't need a summer analyst class if Claude can do the work."
Some of that is real. Some of it is noise. The honest answer is harder than either camp admits.
This post is about three things:
- What's actually shifting in the hiring market, with specifics
- What the scare-stories get right, and what they get wrong
- Five concrete moves you can make this semester
No "the future is bright!" wrapping paper. Just what I'd tell my younger brother.
What's actually changing
Three shifts are happening at once, and most commentary conflates them.
1. The tasks juniors used to do are getting automated.
The entry-level SWE job at a big company used to involve writing boilerplate, triaging bugs, translating PM specs into tickets, and answering your senior's questions. A non-trivial slice of that is now something a senior can do themselves with Copilot in twenty minutes. Same in consulting (first-year analysts used to do the Excel work and the slide formatting), same in law (document review), same in finance (pitch book building).
This doesn't eliminate the job. It compresses it. Where a team used to have six juniors and two seniors, it now has two juniors and two seniors. The team still ships about as much. Sometimes more.
2. Hiring bars have moved, not risen.
Companies aren't looking for "better" candidates. They're looking for different ones. A computer science senior who can ship a working side project using AI tools is preferred over one with a 3.9 GPA and no projects. Three years ago that ratio was flipped. A philosophy major who built a GPT wrapper that got 200 paying customers is beating a business student with a 3.7 from a top-30 school.
This is not a meritocracy suddenly appearing. It's a different definition of "signal." The signal used to be "you followed the prescribed academic path." Now it's "you can actually make a thing that a person would use."
3. The top of the funnel is getting filtered harder.
Every major company now runs AI over applications before a human sees one. Not just ATS keyword matching. Actual language model screening of cover letters for red flags, resume coherence checks, salary-negotiation-history pattern matching. The 400-applications-no-responses story is partly because 380 of those applications were probably auto-rejected in the first five seconds.
If your resume isn't tailored to each role, you're essentially sending bulk mail in 2026. Recruiters' inboxes have the same signal-to-noise problem yours does. They solved it the same way: filters.
What the fear gets right
A few things are genuinely worse than they were in 2022:
- Entry-level tech roles at the biggest companies are harder to get. Not impossible, but the number of new-grad SWE slots at Meta, Amazon, Google combined fell by roughly half between 2022 and 2025. Microsoft and Netflix cut even deeper.
- "Your first two years as a junior" is no longer a reliable training period. If you land one of those roles, you'll be expected to ship production code in your first month, not after six months of shadowing. There's less patience.
- Some fields are genuinely shrinking. Entry-level paralegal work, commodity copywriting, first-line customer support. Not "AI is coming for your job" hand-waving. There are real, measurable headcount cuts in these niches.
If you were planning a career built on being the person who does the boilerplate, that's a reasonable thing to worry about.
What the fear gets wrong
Most of the "AI is taking jobs" discourse fails a basic smell test: the jobs AI is supposedly eliminating still get posted, filled, and paid every month.
The headline layoffs at big tech hide a messier picture:
- New-grad offers at smaller companies are up. Early-stage startups are hiring more juniors than ever because AI tooling makes a smart junior productive faster. The jobs moved, they didn't vanish.
- The roles that require judgment, taste, or customer empathy haven't budged. Designers, PMs, founding engineers, client-facing consultants, field sales. All of these are growing. None of them are automatable with current models, and if the model gets better, the person using it gets more valuable, not less.
- Most people applying for jobs in 2026 are still sending the same 2019-era resume. If you show up with a well-targeted application and proof you can use AI tools effectively, you are competing against a shockingly low bar.
The fear conflates "the job market is harder than it was five years ago" with "there is no job market." Only the first one is true.
Five things you can do this semester
No magic. No hustle-bro nonsense. Five specific moves that work.
1. Build one thing, publicly, using AI visibly
Recruiters and hiring managers are scanning for proof that you can actually use these tools, not just pass a Leetcode question with Copilot turned off. The proof is a project. A working one. With a URL.
It doesn't have to be impressive to a Stanford CS PhD. It has to exist and do something useful for someone. A Chrome extension that summarizes long PDFs. A bot that grades your cover letter. A scraper that tracks apartment listings in your city. Pick something small. Ship it. Put it on your resume with the URL.
When you get the interview, you'll be asked how you built it. The right answer is specific: "I used Claude to sketch the schema, then wrote the auth layer myself because I wanted to understand sessions, then used Cursor to debug the refresh-token edge case." That answer beats "I used AI to build it" by a mile.
2. Tailor every application. Actually every one.
This is the single highest-ROI thing you can do and the one students skip the most.
A tailored resume takes about 10 minutes if you have a base resume set up properly. A generic resume that you paste into 400 job boards takes zero minutes but has about a 0.5% response rate. 40 tailored applications at 10% response beats 400 generic at 0.5%, and the math is not close.
Tailoring means: the job description mentions Kubernetes, so your experience with containers is at the top. The job is at a startup, so your side project is prominent. The job is in healthcare, so the hackathon where you built a hospital triage tool leads. It's not copy-paste; it's triage.
Tools like Zelume (sorry, we had to plug it) do this in one click, but the underlying move is what matters, not the tool.
3. Get fluent with two or three AI tools, not twenty
"AI fluency" has become such an overused phrase it's almost meaningless. Here's the concrete version: pick Claude or ChatGPT, pick Cursor or Copilot, pick one image or data tool, and actually use them in real work every week for a semester. By graduation you should be able to tell a stranger at a bar the exact difference between how Claude and ChatGPT handle a long context window, or why you prefer Cursor over Copilot for certain tasks.
That kind of specificity reads as "I actually use this" instead of "I wrote 'AI' on my resume." Interviewers can tell the difference in about 30 seconds.
4. Own one domain that AI amplifies instead of replaces
There's a list circulating of "AI-proof" jobs. Most of it is wrong. The correct framing is jobs where AI amplifies the best people and makes the average ones less necessary.
Examples:
- Product management. Good PMs are 3x more productive with AI tools (writing specs, synthesizing research, drafting user stories). Mediocre PMs are now redundant.
- Design. Figma AI + your taste is unbeatable. Figma AI + no taste is a mess.
- Data analysis. SQL + Python + Claude lets one analyst do the work of five.
- Sales engineering. AI demos, AI-drafted follow-ups, AI-assisted objection handling. The rep still has to read the room.
Pick one of these, go deep, build proof, and you're in a market where demand is rising, not falling.
5. Interview like a person who uses these tools, not one afraid of them
When the interviewer asks "how do you feel about AI in your industry?", the wrong answer is a political hedge. The right answer is a concrete, opinionated one.
"I use Claude for the first draft of any technical spec, then edit heavily. It's saved me about 4 hours a week. The thing I've noticed is that it's bad at deciding what to build; it's great at writing up what you've already decided. So I spend more time on the deciding."
That's a person who has thought about this. That person gets hired.
The one-paragraph version
The job market is harder than it was. Some of it is AI, some of it is the 2022–24 macro correction, most people are conflating the two. The candidates who are getting offers are the ones who show they can use AI tools effectively, ship real projects, tailor every application, and talk about their work in concrete terms. None of that requires you to be a prodigy. It requires you to do the work that everyone else skips.
See you at the interview.
Notes on ATS parsing, internship pipelines, and the things we built Zelume to solve.
Action Verbs for Software Engineer Resumes (With Real Bullet Examples)
Built, shipped, debugged, scaled. The verbs that work on an engineering resume, the ones that backfire, and wo…
11 Synonyms for 'Managed' on a Resume (and When Each Fits)
'Managed' appears on almost every resume, often three times. Here are 11 more precise alternatives, sorted by…
What to Say Instead of 'Responsible For' on a Resume
'Responsible for' describes your job posting, not your work. Here are 12 stronger replacements, matched to wha…