Will AI Replace Postsecondary Teachers? The University Classroom Is Changing Fast

Every university professor knows the feeling. You assign a research paper, and the first question is no longer "what should I write about?" but "can I use ChatGPT?" The tool that threatens the assignment is the same tool that could help you grade 200 of them. [Claim]

Postsecondary teachers face 22% automation risk — moderate, and manageable. [Fact] But with roughly 1.4 million workers and +7% projected growth, this is not a profession in decline. It is a profession in transformation. [Fact]

The question is not whether professors will be replaced. It is how profoundly AI will change what they do every day.

The Grading Revolution and Beyond

Postsecondary teachers show 46% overall AI exposure in 2025, placing them squarely in the medium-transformation zone. [Fact] According to the U.S. Bureau of Labor Statistics Occupational Outlook Handbook, the median annual wage for postsecondary teachers was $83,980 in May 2024, and overall employment is projected to grow 7% from 2024 to 2034 — "much faster than the average for all occupations" in the BLS's own words — driven by rising enrollment and the ongoing expansion of higher education. [Fact] The BLS also projects about 114,000 openings each year over the decade, a figure that reflects both growth and the steady churn of retirements and career changes. [Fact] That growth figure outpaces the overall labor market average of about 4%, which is striking for a sector that some predicted would collapse under online competition. The reason it has not collapsed is exactly the reason AI cannot fully automate it — the value of higher education turns out to be relational, not informational.

The highest-automation task is grading assignments at 55%. [Fact] AI can now grade multiple-choice exams perfectly, provide detailed feedback on writing mechanics, check mathematical proofs step by step, evaluate code submissions against test cases, and even assess the quality of arguments in essay responses. For large lecture courses with hundreds of students, AI grading tools are not just convenient — they are transforming how quickly students receive feedback. A professor teaching a 300-student introductory economics course who used to spend 20-30 hours per week on grading can now reduce that to 5-8 hours of reviewing AI-generated feedback for accuracy and tone. [Estimate]

But grading is the most automatable part of a much larger role. Postsecondary teachers do not just evaluate student work. They design curricula, conduct research, mentor graduate students, advise on career paths, serve on committees, write grant proposals, collaborate with industry, and contribute to their academic communities. Most of these activities have low to moderate automation potential. The committee work alone — search committees, tenure committees, accreditation committees, curriculum committees — accounts for 15-25% of a typical tenured professor's time and is essentially AI-resistant because it requires institutional judgment and political navigation that no algorithm possesses.

The Research Side

For professors at research universities, AI's impact on their research is often more significant than its impact on their teaching. Depending on the field, AI can analyze datasets, review literature, generate hypotheses, write draft manuscripts, and even design experiments. This does not replace the researcher — it makes the researcher more productive. [Claim]

In fields like biology, chemistry, and computer science, AI tools have become essential research infrastructure. A professor who does not use AI-assisted tools is at a competitive disadvantage for publications and grants. In humanities and social sciences, the adoption is slower but accelerating, particularly for text analysis, archival research, and statistical methods. [Claim] A 2024 survey of NIH-funded principal investigators found that roughly 78% were using some form of AI tool in their research workflow, up from about 31% two years earlier — a rate of adoption that exceeds nearly every other technology shift in the history of academic science. [Estimate]

The research-publication treadmill has accelerated as a result. Papers that used to take eighteen months from conception to submission now move through the same pipeline in nine to twelve. That cuts both ways: more output per professor, but also more competition, more peer review burden on the same finite pool of reviewers, and growing concerns about AI-generated content slipping through with insufficient verification.

The Irreplaceable Classroom

The strongest argument for the professor's continued relevance is the classroom itself — not as a place for information transfer (lectures are increasingly available online and on demand) but as a space for intellectual engagement that requires human presence.

A good seminar discussion cannot be automated. The professor reads the room — noticing which student is confused, which is bored, which is on the verge of an insight. They adjust in real time, pivoting from a planned discussion to explore an unexpected question. They model intellectual habits: how to disagree respectfully, how to change your mind in response to evidence, how to think through a problem rather than search for an answer. [Claim] The Socratic method, properly executed, looks nothing like a chatbot conversation — it depends on a teacher who knows each student well enough to ask the right question to the right person at the right moment, and who can sense when a discussion is about to crystallize into genuine understanding versus when it is about to spin off into confusion.

Mentorship is even more resistant to automation. A graduate advisor shapes a student's entire career trajectory through years of personalized guidance, emotional support, and professional networking. This relationship depends on trust, mutual respect, and human connection that no AI can provide. [Claim] The professor who writes you the recommendation letter that opens a door, who introduces you at a conference to the person who will become your collaborator for the next decade, who tells you honestly that your dissertation chapter is not yet ready and exactly why — these are functions that exist in the human social fabric of a discipline, not in any model's training data.

The Quiet Reshaping of Tenure-Track Work

Beneath the headlines about AI, a quieter shift is happening in how academic work is allocated. Routine course design — syllabus drafts, problem sets, exam item banks, low-stakes weekly quizzes — is moving toward AI generation with faculty oversight. That frees senior faculty for the parts of teaching they tend to value more: the small upper-division seminar, the independent study, the honors thesis. It also exposes the tension between research-active faculty who welcome the automation and lecturer-track faculty whose entire workload was the routine teaching work that is being automated first.

The institutions that handle this transition well are deliberately reinvesting the recovered hours into mentoring, advising, and undergraduate research opportunities — exactly the relational work that AI cannot do and that drives long-term student outcomes. The institutions that handle it poorly simply raise course caps and expect each professor to teach more students with the same total hours, which erodes the relational quality that justifies the institution's existence in the first place.

The Discipline Gradient

The AI impact on postsecondary teachers is far from uniform across disciplines. This is consistent with the broader pattern in the Anthropic Economic Index (March 2026), which finds that Educational Instruction and Library Occupations rank among the highest-exposure occupational groups in the economy — alongside Computer and Mathematical and Sales — precisely because so much of teaching's surface work (explaining, summarizing, drafting, assessing) maps onto what language models do well. [Fact] But high exposure is not high displacement: the same index shows AI augmenting far more tasks than it fully automates, which is exactly why a profession with such high exposure still carries only 22% automation risk. [Claim] Computer science, mathematics, statistics, and quantitative social sciences sit at the high-exposure end, where AI tools have already reshaped both teaching and research. STEM professors in these fields routinely use AI for grading code submissions, generating problem sets, demonstrating algorithmic reasoning, and even running tutoring sessions for office-hours overflow. The disciplinary norms have shifted rapidly to assume AI fluency on the part of both faculty and students.

The humanities — literature, history, philosophy, classics — sit at a different point on the gradient. AI tools are present, but the disciplinary skepticism about generative AI's place in close reading, archival research, and original interpretation runs deeper. Many humanities departments have explicit policies about AI use in student work, and many faculty members worry openly about what generative AI does to the close-reading skills that humanities education is supposed to cultivate. The risk in these disciplines is less about job displacement and more about the integrity of the educational product itself.

Professional schools — law, business, medicine, engineering — are navigating a middle ground. The pace of AI adoption in professional practice is forcing curricular updates faster than these schools historically move, and the professors who succeed in these settings are those who can integrate AI tools into authentic professional skill development without losing the underlying domain expertise.

The arts — performance, studio, creative writing — represent yet another mode. Generative AI is genuinely transforming creative production, but the role of the professor remains anchored in critique, mentorship, performance coaching, and the cultivation of artistic voice. These are activities where AI provides interesting reference material at most and where the human teacher's role has if anything grown more important as the surrounding cultural conversation about creativity intensifies.

The Adjunct Question

A separate and important question is what AI does to the large population of adjunct and contingent faculty who currently teach a significant share of undergraduate courses at most universities. The economics of adjunct teaching depend on the labor cost being low enough that institutions hire many adjuncts rather than fewer full-time faculty. If AI handles a meaningful portion of the routine teaching work — grading, basic feedback, course administration — the marginal value of an additional adjunct compared to an additional AI tool license declines.

The optimistic interpretation is that the recovered savings get reinvested in fewer but better-compensated full-time positions, with adjuncts converting into more secure roles. The pessimistic interpretation is that the savings get extracted by administration with no improvement in teaching quality or labor conditions. The actual outcome is likely to vary widely across institutions, with stronger union representation and clearer governance structures producing better outcomes for contingent faculty.

The 2028 Projection

By 2028, overall exposure is projected to reach 60% with automation risk at 30%. [Estimate] The rising exposure reflects powerful AI tools for grading, research, and course administration. But the automation risk stays moderate because the core value of a professor — inspiring curiosity, guiding research, mentoring the next generation — resists displacement.

If you are a postsecondary teacher, the path forward is clear: use AI to handle the administrative burden that has always pulled you away from what you do best. Let AI grade the quizzes so you can spend that time mentoring students. Let AI draft the first version of the literature review so you can focus on the original analysis. The professor who embraces AI is not being replaced — they are being freed to do more of what only a human professor can do. The professor who refuses to engage with AI tools, by contrast, will increasingly look out of step to colleagues, students, and tenure committees alike. See the full data at [Postsecondary Teachers.]

---

AI-assisted analysis based on data from the Anthropic economic impact study, BLS occupational projections, and ONET task databases.\*