At one real estate firm, new hires are handed a neighborhood and told to walk it. Not read about it, not pull the comps from a database: walk the actual blocks, count the boarded storefronts, notice which corner gets more foot traffic at five in the evening than at noon, and write up what they think the block is worth. Only after they've turned in that judgment do they see what the firm's AI system already produced: a market assessment built in minutes, denser with data than anything a twenty-two-year-old could gather in an afternoon.
The AI report usually wins on comprehensiveness. That was never the point of the exercise. The point is the gap between the two documents, because that gap is where the new hire's judgment gets built, and where the firm can see, in specific and measurable terms, whether it's growing.
This is one small example from a much larger shift that McKinsey documented in July, in research on how companies are training entry-level workers now that AI has absorbed so much of the routine work junior employees used to cut their teeth on. The finding that matters for anyone under twenty-five weighing a next step: judgment, not credentials, is becoming the trait that decides who gets hired and who gets trusted with real problems, and the young people best positioned for that shift are the ones who already understand how they think before they walk in the door.
The anxiety behind that question is real, and it shows up in the numbers. In a recent survey of graduating seniors that McKinsey's researchers cite, pessimism about starting a career climbed from 46 percent to 62 percent in just two years, and most of the pessimists pointed straight at companies hiring fewer entry-level workers as the reason. The unemployment rate for recent college graduates in the US has been climbing since 2019, sitting at roughly 5.7 percent as of early 2026, and roughly four in ten are working jobs that don't require the degree they went into debt for. McKinsey's own reporting flags real disagreement among economists about how much of this traces to AI specifically versus the rise of remote work, which makes it harder to train anyone new from a distance. Whatever the exact cause, the effect on a nineteen-year-old deciding what to study lands the same way: the old, informal path from novice to expert can no longer be assumed.
The headlines about AI and entry-level jobs can make the career ladder sound like it's disappearing. What the McKinsey research actually documents is something more specific: the ladder is being rebuilt around a different mechanism, and the real estate firm's walking exercise shows that mechanism in miniature. Researchers Bryan Hancock and Charlotte Seiler call it the answer-key model. A new hire attempts something on their own, the AI produces its own version, and a manager sits down with both and talks through where they diverge. Bank of America is running a version of this at scale, holding its 2026 intern class at nearly four thousand people while training them through simulations built to compress years of on-the-job judgment into a matter of weeks. The bank's head of global talent has said the goal is to give people the experience they would once have gained doing the routine tasks AI now handles.
This structure works because of what happens during the comparison itself. In a set of clinical studies, doctors who were simply handed an AI tool to help with diagnosis barely improved over time. Doctors who made their own diagnosis first, then reconciled it against the AI's reasoning, got measurably better at diagnosing on their own, reaching roughly the same level as the AI model itself. The reverse held too: when people used AI to complete a technical task they couldn't do unassisted, the ability disappeared the moment the AI was taken away. The attempt has to come first for anything durable to form.
None of this erases the older problem young job seekers already know by heart, the one where every entry-level listing somehow asks for two years of experience. We've written before about how to break that particular catch-22, and the short version still holds: a portfolio and demonstrated work beat a blank resume every time. What's changed is what happens after you get in the door. The old apprenticeship ran on proximity: you sat near someone experienced, did the grunt work, and picked up their instincts by watching them make calls you weren't yet trusted to make. AI has absorbed a lot of that grunt work, so the proximity has to be built on purpose now, through structures like the answer-key model.
The question that changed
For decades, the shorthand question in hiring was simple: what did you study? McKinsey's survey data on entry-level hiring shows that question losing ground fast. When executives rank the traits they actually look for in new hires now, adaptability, problem-solving, resilience, and the ability to reason through something unfamiliar sit at the top of the list, ahead of narrow domain training. One head of HR at a Fortune 100 financial firm described the hire they're now looking for as a general athlete: someone with strong learning instincts, decent relational skills, and a working familiarity with AI.
That widening extends further than most people expect. Because knowledge is more accessible than it used to be, and because a well-built AI system can hand a newcomer a huge amount of institutional context on day one, companies are also opening entry-level roles to what the nonprofit Opportunity@Work calls STARs, people skilled through alternative routes who never finished a four-year degree but built real capability through work. Roles themselves are being drawn wider too. Some companies are hiring one person into a commercial role that spans sales, marketing, and strategy, on the logic that a generalist with strong judgment can be pointed at whatever problem shows up that week.
If you're seventeen or twenty-two and you don't have a fixed target field yet, this is, if anything, good news. A settled answer to "what do you want to do" was never the real starting point. Curiosity and a willingness to test an idea against reality count for more, and the hiring data backs that up: employers describe optimizing for how a person thinks and adapts, with the age at which someone locked onto a lane barely registering.
What a career assessment is actually for
Here is the practical version of all this research. If judgment is the trait employers are now screening for, and judgment is built fastest through an honest attempt followed by real feedback, then the young person with an advantage walks into that first assignment already knowing something about their own reasoning: what kind of problems hold their attention, what motivates them to push through the hard part, what personality traits show up under pressure, and what working environment lets their best thinking surface. That's a different question than "what job pays well" or "what's in demand this year," and it's the question a structured career assessment is built to answer.
A good assessment works the same way the answer-key model does. You answer honestly first, which is your attempt. What comes back is a reflection of that attempt, measured against a wider set of data than any one person could hold in their head, in CareerMatch's case a database of 4,000+ researched roles, weighed against how you think, what motivates you, and the kind of environment that brings out your best work. Treat the result the way the McKinsey research treats every AI comparison: as a starting point worth testing against real experience, a summer project, a shadow day, a conversation with someone actually doing the job, before deciding the assessment got it right.
There's a detail in the McKinsey research worth sitting with here. The gap between a junior employee's independent attempt and the AI's output is something a company can watch over time, and a gap that narrows is treated as real evidence that judgment is forming, a signal the old apprenticeship model was never able to measure so directly. Apply that same habit to yourself. Every time you test an assessment result against something real, an internship, a class project, a conversation with someone in the field, you're generating your own version of that narrowing gap, and it's worth paying close attention to what closes and what doesn't.
What this means if you're the one paying tuition
Parents reading headlines about AI and entry-level jobs often reach for the safest-sounding answer: push for a technical major, a credential the machine supposedly can't touch. The McKinsey research points somewhere else. The traits holding up best under this shift are broad ones: adaptability, reasoning, resilience, and comfort working alongside AI tools. These transfer across fields in a way a single narrow credential rarely does. Technical skill still matters, but on its own, without the judgment to apply it, it has become a weaker bet than it used to be.
There's also a fair question buried in the headline of the McKinsey piece itself: who actually trains the next generation, if AI is doing the work a manager used to hand a junior person to learn on? Some of the researchers studying this point to medicine for an answer, specifically the residency model, where a new doctor practices under a designated senior physician before earning the right to work alone. A version of that idea, sometimes called a preceptor model, is starting to show up in how technology and finance firms structure junior roles: a senior person formally mentors a small group of newcomers, working alongside AI tools together, so the senior can watch exactly what the junior accepts, questions, or gets wrong. The mentor's job shifts from answering questions to teaching judgment directly.
That's worth sitting with before the next conversation about which major to lock in. CareerMatch built a dedicated resource for exactly this conversation, because the research a parent needs to have that discussion well is different from the research their kid needs to choose a direction, and both matter more now than they did five years ago.
Back at that real estate firm, the new hires who improve fastest are the ones who go back out and walk the block again with a sharper eye, because they finally know which details they missed the first time and why. That's the shape of the next few years of work: knowing exactly what you personally bring to a comparison the machine can run faster than you ever will. Start there, with an honest look at how you think, before the job market asks you the same question.