Why Some Early Career Workers Thrive With AI While Others Fall Behind

New research in Harvard Business Review helps explain why two early-career workers with nearly identical resumes can get wildly different results from the same AI tools. In a field study of 523 early-career professionals run by KPMG and the University of Texas at Austin, only 50.1 percent outperformed AI working on its own, and that gap had almost nothing to do with how smart or credentialed anyone was. At Metaintro, we read studies like this because they tell job seekers exactly which skills employers will reward next. Here is what separates the workers who thrive with AI from the ones who quietly fall behind.

What did the research actually find?

The study, published in Harvard Business Review on July 23, 2026, sorted its 523 participants into three groups after they completed business tasks alongside an AI agent built for the work. The labels are worth memorizing because they describe three very different futures. The researchers called them AI amplifiers, AI delegators, and AI apprentices.

AI amplifiers made up 50.1 percent of the group and outperformed the AI-only baseline by orchestrating the workflow, framing the problem strategically, and refining results across several rounds. AI delegators, about 25.8 percent, produced work that was roughly as good as the AI alone because they accepted its output with little scrutiny and added little of their own. AI apprentices, the remaining 24.1 percent, actually landed below the AI baseline, meaning they made the final product worse than if they had simply handed the task to the machine.

Read that last number again, because it is the uncomfortable heart of the study. One in four early-career workers in a controlled setting dragged down output that AI could have delivered on its own. In a real job, that is the person whose manager slowly stops trusting them with the interesting work. Understanding why that happens is the first step to making sure it is not you, and it connects directly to the productivity bar AI has raised for new hires across knowledge work.

Why do foundational skills fail to predict who thrives?

Here is the finding that should reset how you think about your own career. The people who fell behind were not less capable on paper. According to HR Dive's reporting on the study, the underperformers scored just as high on critical thinking, domain knowledge, and AI literacy as the amplifiers who pulled ahead, and the primary differentiator was how they worked with AI rather than what they knew. The Harvard Business Review authors went further, noting that the apprentices who ended up below the baseline actually scored higher than the delegators on every foundational skill measured, yet still finished last.

That result breaks the story most of us were told about school and work. For decades the path was clear, which was to accumulate knowledge, prove it on a test or a resume, and let credentials open doors. The research suggests that AI has quietly changed the rules, and that raw knowledge is now the price of entry rather than the edge. This is the same shift we described in our look at the 2026 AI acumen gap that bosses are testing for, where fluency with AI matters as much as the substance behind it.

It also explains a strange paradox many new grads feel right now. You can be genuinely talented, hardworking, and well trained, and still watch your output look average once AI enters the picture. That is not a personal failing, it is a signal that a different muscle is being tested. We unpacked the deeper version of this tension in our piece on the AI paradox where replacing experts still requires people who can learn, and the study puts hard numbers behind it.

What separates an AI amplifier from an AI delegator?

The difference comes down to a single relationship, which is how you treat the AI. As HR Dive summarized it, the high performers treated AI as a collaborator that needed direction, oversight, and judgment, rather than a vending machine that spits out finished work. Delegators did the opposite, taking the first answer at face value and moving on.

Picture two analysts asked to summarize a messy market report. The delegator pastes the report into an AI tool, copies the summary, tidies a sentence or two, and sends it. The amplifier starts by telling the AI what decision the summary needs to support, asks it to flag the three weakest assumptions, pushes back when a claim looks thin, and runs the draft again with sharper framing. Same tool, same starting knowledge, completely different output. That habit of steering and interrogating is the skill the study rewards, and it is the core of what we describe in our guide to the AI skills that help you thrive at work.

What makes this encouraging is that steering is a practice, not a personality trait. Ashish Agarwal, a professor at the University of Texas at Austin who worked on the study, said the team wanted to understand what enables some individuals to consistently create value beyond what AI can produce on its own. Rahsaan Shears, an AI enterprise transformation leader at KPMG US, put the takeaway even more plainly, saying that how people applied their knowledge and skill is what made the difference, and that the gap is coachable. Coachable is the key word for anyone early in their career, because it means the advantage is available to you no matter where you started.

Why are entry-level jobs the hardest hit by this shift?

The reason this research lands so hard on early-career workers is that AI is eating exactly the kind of work juniors used to learn from. The Harvard Business Review authors noted that many entry-level tasks that once served as learning opportunities are increasingly being delegated to AI-powered workflows, which are now capable of handling analytical, information-heavy assignments. When the first draft, the basic research, and the routine analysis all get automated, the rungs on the ladder that used to build expertise start to disappear.

That creates a genuine trap. If the tasks that once trained new hires are gone, how do you build the judgment the study says matters most? We have written about this squeeze before in our explainer on what AI coming for entry-level jobs means for new grads, and the KPMG data sharpens the warning. The workers who thrive are not the ones avoiding AI to protect their learning, they are the ones using AI while deliberately keeping their own judgment in the loop.

There is a second, quieter risk hiding here, which is skill erosion. When AI handles everything, it is easy to stop practicing the reasoning that makes you valuable. We called this the comfort trap in our piece on automation eroding the skills that keep you hireable. The delegators in the study were not lazy, they were comfortable, and comfort with a tool that is improving faster than you are is a slow way to fall behind.

What does this mean for your career right now?

The practical message is that your competitive edge has moved. It used to sit in what you knew, and now it sits in how well you can direct, question, and improve the work of an AI that already knows a great deal. That is good news if you are early in your career and worried about competing with more experienced colleagues, because seniority is not what the study rewarded. Judgment applied to AI output is, and that is something you can start building today.

It also changes how you should present yourself to employers. Hiring is already shifting toward specific, demonstrable skills rather than titles or years served, a trend we tracked in our report on employers chasing specific skills in 2026. Being able to show a hiring manager how you turned a mediocre AI draft into a strong final product is now a more powerful story than a line about being detail oriented. Concrete examples of you steering AI toward a better outcome are the new proof of competence.

None of this means chasing every tool or certificate. The research is clear that AI literacy alone did not separate the winners from the losers, so stacking credentials is not the answer. What matters is the applied habit of collaboration, which is why our advice in the guide to five skills that beat a job title in AI workplaces focuses on behaviors you can practice rather than badges you can collect.

Which AI skills should early-career workers build first?

Start with framing. Before you ask an AI for anything, get specific about the decision the work needs to serve and the standard a good answer has to meet. Amplifiers in the study won partly by framing problems strategically before touching the tool, and that front-end thinking is where most of the value is created. A vague prompt produces a vague draft that a delegator will accept and an amplifier will reject.

Next, build the habit of structured critique. The apprentices in the study offered plenty of criticism, but their critiques often chased irrelevant issues or steered the AI in the wrong direction, so the feedback made things worse rather than better. The skill is not simply doubting the AI, it is knowing which parts of an output actually matter and pressing on those. That is a learnable discipline, and our structured plan to build AI skills breaks it into a routine you can follow week by week.

Finally, treat iteration as the job rather than a nice-to-have. Amplifiers refined their results across multiple rounds instead of shipping the first draft. That means budgeting time to run a task two or three times, comparing the versions, and keeping the strongest pieces of each. These agentic collaboration habits are quickly becoming the baseline expectation for hiring, a shift we covered in our analysis of the career skills defining the next wave of hiring. Combined, framing, critique, and iteration are the practical toolkit the study rewards, and they will keep you employable through constant change.

How can you tell if you are amplifying or just delegating?

Run a simple honesty check on your own week. When you used AI, did you tell it what decision the work had to support, or did you just describe the task? Did you push back on anything it produced, or did you accept the first answer because it looked polished? Did you run the work more than once, or did you ship version one? Delegators are not careless people, they are busy people who let a confident-sounding output end the conversation too early. The tell is speed with no second look, because the moment an answer sounds authoritative it is easy to assume it is also correct, and that assumption is exactly what the study punished. A short pause to interrogate the work is what turns a rushed delegator into a deliberate amplifier.

If your answers lean toward accepting and shipping, you are likely delegating, and the fix is not to work harder but to work more deliberately. Add one question to every AI interaction, which is what would make this wrong, and force yourself to answer it before you use the output. That single habit moves you toward the amplifier behavior the study measured, and it compounds fast. As you build it, pair it with a structured job search so your new edge actually reaches employers, using something like our AI-assisted job search plan to turn the skill into offers.

The larger point is that the age of AI has not made early-career talent less valuable, it has changed what talent looks like. The workers who thrive are not the ones with the best grades or the most tools, they are the ones who stay in the loop, keep their judgment sharp, and treat AI as a partner they are responsible for directing. At Metaintro, we believe that shift is an opportunity, because the edge it rewards is one anyone can choose to build.

Related Articles

• The 2026 entry-level squeeze and the productivity bar AI raised for new hires
• What AI coming for entry-level jobs means for new grads
• The 2026 AI acumen gap bosses are testing for
• The AI adoption gap and employer training failure
• Why most workers cannot connect AI training to their job
• How AI certifications boost salary and career growth
• The critical skills your team is losing to AI deskilling
• Why employers are chasing specific skills in 2026
• Why the 2026 career edge is relationships, not AI
• The AI skills required for promotion at Accenture
• The older-worker AI advantage

People Also Asked

Q: What is the difference between an AI amplifier, an AI delegator, and an AI apprentice?

A: In the KPMG and University of Texas study, AI amplifiers, who were 50.1 percent of participants, outperformed AI alone by directing and refining its work. AI delegators, 25.8 percent, matched AI by accepting its output with little scrutiny. AI apprentices, 24.1 percent, fell below the AI baseline because their critiques often steered the work in the wrong direction.

Q: Do AI skills matter more than intelligence or credentials for early-career workers?

A: The study suggests behavior matters more than raw ability. Per HR Dive, underperformers scored just as high on critical thinking, domain knowledge, and AI literacy as top performers, so the differentiator was how they worked with AI, not what they knew. You can read more in our guide to the skills that beat a job title.

Q: Can you learn to work better with AI, or is it a fixed trait?

A: It is learnable. A KPMG leader who worked on the study said how people applied their knowledge is what made the difference and that the gap is coachable. Building the habits of framing, critique, and iteration through a routine like our structured plan to build AI skills is how you close it.

The job market rewards people who can see where it is heading. AI is not erasing early-career talent, it is rewarding a new kind of it, and Metaintro exists to help you build that edge and put it in front of employers who are hiring for it. Create your free Metaintro profile to stay ahead of the shift and match with verified roles that value the skills you are building.