After generative AI rapidly penetrates the workplace, the first chill may not be felt by veteran workers who have already cemented their positions, but by newcomers who are just about to step into the job market. From the old career path of “start with entry-level work and then slowly work your way up,” to today’s trend where companies more often seek people with experience who can hit the ground running, AI is rewriting not only job duties, but also the “training ground” that young people once relied on to grow.
In “Business Talk, No Nonsense,” the hosts—Linga, along with Bradley—delve into this phenomenon. Bradley has an NTU medical degree, a Harvard public health background, and a McKinsey background, and has served as a senior executive at 乐天医药 and Appier. The two try to answer a question that has become increasingly uncomfortable: Will AI not first replace senior employees, but instead prevent new hires from even getting a chance to step on stage?
After finding Harvard- and Stanford-related research, the program discovered that after ChatGPT was released, entry-level job openings indeed showed a clear decline—especially in job categories with higher exposure to AI. Jobs such as administration, secretarial work, entry-level sales, marketing, and similar roles were hit particularly directly.
AI plus senior colleagues is enough—why train new hires?
After citing Harvard- and Stanford-related research, the program found that after ChatGPT was released, entry-level job openings indeed showed a clear decline—especially in job categories with higher exposure to AI. Jobs such as administration, secretarial work, entry-level sales, marketing, and similar roles were hit particularly directly. This means companies are not completely stopping hiring; instead, they’ve begun to reassess: Since a slightly experienced employee paired with an AI tool can produce output that previously took two or three people, is it still necessary to spend the same budget to train multiple new hires?
Bradley said bluntly that this shift is already quite obvious in day-to-day corporate operations. In the past, companies were willing to give new hires half a year or a year to do training, rotation, and learn the industry. But now that kind of patience is shrinking fast. For managers, the numbers have become painfully straightforward: should they hire two or three entry-level employees and also arrange for senior colleagues to mentor them while investing training resources? Or should they simply hire one experienced person and pair them with the best AI tools to go straight onto the battlefield? In most companies that prioritize efficiency and immediate effectiveness, the answer is practically self-evident.
So the disappearance of entry-level roles isn’t just as simple as “a few fewer job opportunities.” It’s that the entire learning structure is starting to loosen.
Because many tasks that new hires used to practice with are exactly the kinds of work AI is best at handling: finding information, organizing summaries, translating, writing basic code, and doing preliminary analysis. These jobs were previously repetitive and tedious, but they were the starting point for new hires to build familiarity, understand quality standards, and observe how senior people judge situations. Now these tasks can be completed by AI in just a few minutes. New hires don’t just lose a practice ground—they also get fewer opportunities to sit beside and watch how managers revise things, how they think, and how they make judgments.
In the AI generation, do people really know what “good” looks like?
Linga also mentioned that this shift is already reflected in the ability gap between students and early professionals. For the past generation without AI tools, making reports meant starting from scratch—finding sources, verifying them, building hypotheses, and going from a blank page all the way to a presentation they could actually take onstage. Although these people may have been slower, they usually understood better how to go from 0 to 100.
In contrast, this new batch of students has already lived in a world of ChatGPT, Grok, and Gemini since university. They’re used to throwing the same question at different AIs and then quickly stitching the outputs into a slide deck or spreadsheet. It may look more efficient, but it also makes it easier to hand in the AI’s 60 or 80-point answer without the judgment that’s needed.
(If AI can do an 80, then people who can’t reach 100 are destined to be eliminated! McKinsey and Harvard alumni advise new hires to do it this way)
This also puts managers into a new dilemma. Bradley gave an example: in consulting work, in the past you might ask a junior to write an email requesting information from a client. It looks like it’s only writing text, but in reality it’s an important process for training the other side to understand the client’s context, to grasp the communication rhythm, and to judge word choice and degree. But now, the situation often becomes: once a manager sees it, they can tell immediately whether the email isn’t mature enough, or whether it’s clearly AI-generated and still a distance away from being usable.
Advice for young people: proactively understand the standards in your manager’s head
However, the two didn’t stop at anxiety itself; they went further to discuss how, as the first step disappears, young people can build a new training ground for themselves.
Linga believes the first thing is to proactively understand the standards in your manager’s head. Because many workplace experts don’t simply do better because they work harder; they have a whole set of judgment checklists in their minds: whether sources should be cross-validated, whether a researcher’s background should be checked, whether conclusions should explicitly state limitations, and how to structure the logic of a presentation. In the past, these standards may have been hidden in repeated revisions and verbal guidance.
But in the AI era, many of these have already been turned by managers into prompts and constraints. For young workers, what really matters isn’t whether you can follow the tools—it’s whether you have a way to first figure out what “good” actually means.
Structure and judgment are the parts AI can’t replace
The second thing is, in an age of information overload, to retrain your reading and synthesis abilities. AI can scan 25 reports at once and quickly give you a summary—but that also makes people easily assume they already understand the problem. Linga believes the more this era accelerates, the more you have to force yourself to pick out truly high-quality content, spend time digging in, and look at how the author defines the problem, breaks down the industry, designs the approach, and validates the hypotheses—and ultimately how they make judgments based on experience.
These paths of structure and judgment are the parts AI can’t directly internalize for you.
Bradley also reminds young workers that if they want to move closer to higher-level work, the first step is always “buy entry with results.” It’s not just about completing tasks—it’s about doing enough that your manager can feel confident and be willing to hand you more complex things. Beyond that, you also need to learn to proactively cover gaps. Because many higher-level jobs in the workplace—closer to the decision core—aren’t without opportunities; they’re just that no one is doing them.
For example, a task that superficially looks like analyzing conversion rates: the real key often isn’t the numbers themselves, but what decisions the analysis is meant to support. If your manager wants to know which customer segment the next marketing budget should allocate more toward, then you shouldn’t just turn in conversion rates; you should take one step further and fill in the customer acquisition cost, Lifetime Value, ROI—and even propose decision language like “If the budget increases by 20%, how much might revenue increase.” The ability to turn analysis into recommendations is one of the most valuable things in the AI era.
What kind of people do future companies want: moving forward in ambiguity, knowing what counts as good
When talking about what kind of people companies should be looking for, Bradley also gave a very clear answer. First, it’s people who can move forward in ambiguity. Because in the AI era things change too fast—three months from now, your job content might be completely different. If someone has to wait until every detail is defined clearly before they act, not only will they suffer themselves, it also means their job content is likely to be replaced by AI sooner or later.
Second, can they judge AI’s output? Do they know what kind of results count as “good”? “Being able to use AI” is already a baseline requirement now. What really creates the difference is whether, after reading an AI-generated market analysis, you have your own viewpoint. Do you know which parts need to be verified? Which parts must not be accepted at face value? Because data is not rare anymore today—the real scarce resource is perspective.
Entry-level job hunting is getting harder for new hires, and entry-level roles are starting to loosen
Third, can they self-upgrade? Faced with a fast-changing environment, companies would rather have someone who doesn’t have as many skills right now but learns quickly than someone who has a lot of skills yet stays stuck in place.
That’s why the two both believe that today’s entry-level openings are less about “disappearing” and more about being redefined. Companies don’t mean they no longer need young people; they just no longer only want people who can do repetitive tasks. The new starting line becomes whether you understand AI, whether you have viewpoints, whether you can quickly take initiative to learn, and whether you can turn tools into your own leverage. For many companies, young people aren’t just the ones being guided anymore—they’re also becoming people who bring AI intuition, tool habits, and the rhythm of the new world into the organization.
At the end, the program also offers a quite practical piece of advice: if you’re job hunting right now, the most effective preparation might not be going back to get another AI degree. Instead, build an AI side project for the industry you want to enter. Suppose you want to enter Google’s marketing department—then try to use AI tools to run through a full marketing workflow end-to-end: from identifying problems, designing solutions, to actually demoing results. That way, in interviews you won’t just talk about “I’m very interested in AI.” You can present a real end-to-end project and prove to the interviewer that you can use tools to solve real problems.
For new graduates who are dealing with anxiety about job hunting, the harshest reality may be this: the career staircase that once seemed like a given is truly being dismantled in part by AI.
But as Bradley said, maybe a career has never only been about climbing stairs—it’s more like swimming. The old path was arranged for you; now you have to find direction yourself and swim forward on your own. When the first step becomes less stable, what matters most may not be whether that staircase is still there, but whether you have the ability to proactively find the next move that will help you stay afloat.
This article, “Graduation job-hunting anxiety? Research shows AI directly affects entry-level roles; McKinsey consultant advises new hires to do this,” first appeared on Chain News ABMedia.
