
What Graduates Are Locked Out Of
Companies that used to hire fifty graduate analysts are hiring five plus a stack of AI tools. Companies that used to hire ten junior copywriters are hiring two senior writers and a content model. Law firms that hired paralegal cohorts to grow into associates are running document review through AI and hiring fewer paralegals.
What is being taken over is not the job title; it is the bundle of routine production work the role was built around. Until recently, that bundle was the training ground. New graduates produced the volume, managers corrected it, and judgment came from the repetition. When AI produces that volume at almost no cost, it is the training itself that shrinks, not just the headcount.
The result is a graduate market that looks roughly the same in posting volume but very different in what those postings reward. The roles that still hire well are the ones where AI is either the product, the impossible-to-delegate, or the orchestrated. The roles that have softened are the ones where AI is doing the work and the graduate is doing the rubber-stamp.
Why This Looks Different From Past Tech Shifts
Past technology shifts hit specific industries. Manufacturing automation in the 1980s changed factory floors. Internet adoption in the 2000s changed retail and media. AI in the 2020s is doing something different in kind. It is not replacing roles inside a few specific industries; it is taking over the same layer of routine production work across almost every knowledge industry at once.
That shared layer happens to be where most graduates are hired. The result is a labour-market shift concentrated at the early-career end of the ladder rather than evenly distributed across seniority. Mid-career professionals are seeing their job change. Graduates are seeing their entry point disappear.
Past shifts also tended to open new ways in within a year or two of closing the old ones. Web developers replaced typesetters; e-commerce specialists replaced retail clerks. The new ways in from this shift are visible already, but they sit in fewer companies, ask for more specific skills at the start, and reward planning at the moment of the first job, not after it.
Hardest-Hit Entry Roles and Safer Entry Pathways
Drawn from the full 1592-role index. Median annual wages are US figures from the Bureau of Labor Statistics OEWS survey. The split below is about how the work is built, not a prediction: it tracks where AI takes over the daily work fastest versus where the work cannot be taken over at all.
Browse the full 1592-role index for exposure readings on every tracked role.
Three Ways to Land a First Job in This Market
The graduate market has not collapsed; it has split in two. Three clear pathways are still hiring well. Each asks for a different set of skills, signals, and early decisions. Most graduates land in one of them; the ones who land badly are usually the ones who never chose.
What to Build Before You Send Your CV
Hiring managers in 2026 are not impressed by AI certifications, GPA, or generic project lists. The four things below carry weight because they each demonstrate something AI cannot generate: judgment, context, and accountability for an outcome.
One AI tool used in anger
Pick one category (an LLM, a coding copilot, an agent framework, an image model) and use it on a real task long enough to know its failure modes, not just its features. Hiring managers in 2026 can tell within five minutes whether an applicant has used the tool or only watched a demo.
One artefact AI cannot produce alone
Ship one project, repo, post, or case study where the value is in your judgment about the AI output, not in the output itself. A landing page that performs, a report that changed a decision, an evaluation that flagged a real issue. Output AI generated unsupervised is no longer a portfolio.
One commercial or domain context
Tie your skills to a field, market, or industry. Knowing AI is now the baseline; what sets you apart is knowing AI plus a real field you understand. The graduate who can talk about a specific market is more hireable than the one who can only talk about AI in general.
One reference who can vouch for judgment
An internship, side project, or volunteering stint where someone watched you decide, not just watched you produce. Production output is now AI's contribution. Judgment is yours, and it has to be witnessed by someone who can speak to it.
How to Read a First Offer in This Market
The right entry-level role is not the highest-paid one or the most prestigious one. It is the one whose week-three workflow puts you in the seat where decisions get made, even if the decisions are small. Three questions to ask before signing.
If the answer is the AI or a manager reviewing AI output you forwarded, the role is rubber-stamping. If the answer is you, even at small scope, the role is building judgment.
Roles where you have a named customer, a clear outcome, or a specific commercial number to move hold up well. Roles where the customer is vague and the goal is just throughput do not.
Every modern role uses AI somewhere. The question is whether you direct it (holds up well) or just use what it produces (losing ground). Ask directly; the answer about week three will tell you whether the job builds on itself.
For the full method, see the AI Career Index methodology.
AI Career Snapshot
A free overview that gives you a baseline read on where you stand and the main risk signals. You get your AI Career Index score, your seniority level, your AI exposure level, and the key things driving them.
Run Free Snapshot →AI Career Positioning Report
A full dashboard with a 24-month forecast (Drift vs Strategic Path), deeper exposure detail, and a 6-month Acceleration Plan.
View Pricing →Common Questions From Graduates and Career Starters
The Bottom Rung Has Moved. Move With It.
Graduates entering the workforce now are doing so while the starting point of the career ladder is being rewritten. The companies, roles, and ways in that defined the bottom rung for the last twenty years are not the ones that will define it for the next twenty. That is a generational disadvantage if you treat it as background noise, and a generational advantage if you use it to choose your first job on purpose. Run the snapshot, read your offers carefully, and start at a rung that builds on itself.
Go deeper on every step
9 long-form articles, grouped by which of the four steps above they expand on. Click a step to see its articles.
"I usually dismiss career tools. This one is different, because the results are consistent and it shows how it reaches them. My test for anything is simple: would I trust it enough to make a real decision on it? Here the answer was yes."
"I work in a field that AI is squeezing hard. The report did not pretend otherwise. What it did was show me which part of my role is safe and how to move more of my work into it over the next six months. That honesty is what I needed."
"I had my whole team of twelve take it. The group results flagged two roles that were quietly falling behind and at risk of losing ground to competitors. We fixed both before it showed up in their numbers. At forty-nine dollars a person, it is the best-value spend I have made all year."


