The Class of ChatGPT: What the First AI-Native Graduates Mean for Employers
Editor’s note: Kristi Girdharry is an associate teaching professor of English and the director of the Writing Center at Babson College. Her work explores the impact of AI on writing and research as a co-lead of The Generator, Babson’s interdisciplinary AI lab.
Graduates of the Class of 2026 are the first to enter the workforce after completing their entire college education in the era of generative AI. They arrived on campus months before ChatGPT launched, and they graduated at a moment when AI is embedded across industries.
What that means for hiring is more nuanced than most managers realize.
Over four years, Associate Professor Stephen McElroy, the director of the Writing Program at Babson College, and I conducted more than 140 interviews with Babson undergraduates, meeting the same students once a semester from first year through graduation. We did not set out to study AI, but, as generative AI reshaped classrooms and internships, the project became a view of how students’ judgment around a technology evolved as it moves from novelty to infrastructure.
Employers are already reporting similar shifts. Surveys from organizations such as the World Economic Forum and LinkedIn suggest that while AI use is becoming a baseline expectation, the skills that differentiate employees are increasingly human: judgment, critical thinking, and the ability to work effectively with new tools.
Our interviews show how those capacities develop in practice. Here are some instructive examples from our observations.
Where the Real Shift Happened
In early interviews, students described AI in broad, functional terms: helpful for summarizing readings, brainstorming ideas, or speeding up research. By junior and senior year—particularly after internships—the language changed. AI was no longer a convenience. It was embedded in workflow.
One student talked about interning at a major e-commerce company where AI proficiency was formally evaluated. Her manager had a specific performance review question asking whether interns were “very advanced in using AI.” There were also training sessions on ChatGPT and Google Gemini. Daily integration was expected.
The contrast with classrooms was structural. Professional environments were operationalizing AI. Many academic environments were still negotiating norms. Students adapted. In classrooms, they were cautious. In internships, they were pragmatic. When they returned, the gap felt wider.
For managers, the takeaway lies in where learning accelerates. The sharpest AI maturation occurred where stakes were real: client deliverables, performance reviews, investment decisions. Many of the deepest learning moments still happened in classrooms. What changes the pace is consequence.
From Fluency to Discernment
The most important development across four years was discernment.
Consider one student’s trajectory. As a first-year, he told us he took pride in doing his own work and was “very against” AI. By sophomore year, a venture capital internship had reversed his position. He watched AI used for deal sourcing and started describing it as “a tool now … almost as a search engine that just formats into the correct way.” But he deliberately kept AI out of his writing: “I want to have my own thoughts expressed, and AI isn’t that.”
By junior year, studying abroad, he noticed something in himself: “Sometimes it’s like such an easy instinct to go to AI.” He pushed back against that instinct, choosing independent work to maintain confidence in his own abilities. By senior year, he estimated that 80 percent of his classmates used AI even in classes where it was banned, and he agreed with his professors that AI could diminish critical thinking while still using it himself for research and outlining. He participated in a hackathon at our school and built a website in an hour without knowing how to code.
That is a four-year arc of resistance, adoption, restraint, and calibration. That oscillation, rather than a steady increase in use, is what produced judgment.
Across our interviews, the pattern held: as stakes increased, AI shifted from convenience to calibration. Students became less concerned with speed and more attentive to boundaries in terms of what to delegate and what to own.
Building Ethical Frameworks
Perhaps most striking was how much ethical reasoning students developed on their own. Multiple students, independently, arrived at the same framework: AI should execute tasks that humans define; it should not replace human thinking.
For example, a student who ran his own company used AI daily for business tasks such as drafting emails, optimizing delivery routes, and generating ad copy. He built a custom chatbot by feeding it an entrepreneur’s books and YouTube videos. But he drew a hard line at academic writing and was candid about the cost: “I do almost wonder if it’s lowered my critical thinking abilities.” In a later conversation, he described finding his footing: “I started to realize over the next few months after that, that I can think myself.”
Contrary to popular narratives, these students built internal rules. They distinguished between using AI to deepen understanding and using it to bypass effort. They converged on shared categories: drafting partner or ghostwriter, accelerator or replacement, thinking aid or shortcut. Some of this reasoning came from faculty guidance. Much of it was refined through experience, especially where real stakes forced them to draw lines.
The Opportunity for Employers
Traditional hiring signals were designed for a pre-AI world. Today, they are less reliable. As AI tools make it easier to produce polished work, they are also making it harder to tell who actually understands the work behind it.
One student in our study described classmates who relied so heavily on AI that, without it, they were “debilitated.” He could produce polished, AI-assisted essays in minutes and earn top grades—not because he had mastered the material, but because the assignments did not require him to demonstrate that mastery. If AI can simulate competence in a classroom, it can do the same in hiring.
The question for managers is whether their evaluation methods can still distinguish between production and understanding or reward the ability to simulate it.
Much of what these students learned about AI came from internships and early work environments. The norms modeled by managers—what AI is for, how its output is evaluated, when human judgment must remain central—shaped their habits. That places employers in a position of influence. If traditional signals are weakening, organizations have to build new ones. Three practices matter:
- Make expectations explicit.
- Evaluate process, not just output.
- Reward judgment.
The Class of ChatGPT is not defined by whether they used AI. They almost certainly did. They are defined by whether they learned to use it with discernment.
Latest Stories
Posted in Community
Posted in Community, Entrepreneurial Leadership, Insights
