The Agentic Professor: Exploring GenAI-Supported Futures in Higher Education

For centuries, higher education institutions have occupied a uniquely stable position in society. While technologies, from printing presses to the internet, have continually reshaped how colleges and universities operate, academic structures have remained remarkably resistant to transformation. Professors teach, evaluate, mentor, and certify; students study and progress through academic terms; credentials are awarded through processes deeply embedded in tradition. Today, agentic artificial intelligence (AI) is challenging that stability. In just a few years, AI has evolved from a novelty into a ubiquitous partner in writing, coding, designing, analyzing, simulating, and, increasingly, teaching. Capabilities such as reasoning, multimodal understanding, persistent memory, planning, tool use, and autonomous task execution allow generative artificial intelligence (GenAI) to function less as a reactive tool and more as a proactive cognitive collaborator.

One possible manifestation of this shift is the Agentic Professor: a long-horizon AI instructional partner that can support students across courses, disciplines, and years of study. Regardless of whether such systems emerge in precisely this form, the concept of an Agentic Professor surfaces questions about pedagogy, assessment, governance, and the educational roles that should remain distinctly human.

While AI systems are improving rapidly, the pace and trajectory remain uncertain. Progress in reasoning, pedagogy, and conceptual explanation has been slower and less consistent than progress in domains such as coding, where measured performance has doubled roughly every six to nine months. Still, even a deliberately conservative assumption—capabilities relevant to learning systems doubling approximately every eighteen months—would imply six to seven doublings, or roughly a one-hundred-fold increase, over ten years. The ten-year horizon is illustrative rather than an estimate of when the capabilities described below might emerge; they may well arrive considerably sooner. The point is not to predict a specific outcome but to recognize the scale of change that continued gains could produce and the new instructional models they might enable.

If AI systems become increasingly capable of guiding learning, evaluating complex work, supporting students over time, and drawing on a persistent understanding of each learner, the traditional organization of teaching is no longer the only plausible model. In that context, the Agentic Professor becomes a useful thought experiment for exploring how instruction, assessment, and faculty roles might evolve.

Three principles frame our thinking about the future:

1. Technological capabilities often advance faster than institutions can comfortably adapt.
2. Forecasts frequently fall victim to the static-future fallacy, projecting one change while holding everything else constant.
3. Technologies that challenge perceptions of human uniqueness often provoke resistance beyond purely analytic disagreement.

These observations are important because colleges and universities are not merely instructional delivery systems. For example, early predictions about digital libraries famously assumed that once books were online, libraries would become obsolete, overlooking their roles as study spaces, community hubs, and intellectual commons. A similar error would be to assume that the arrival of Agentic Professors determines the future social fabric of the university in a single direction. If Agentic Professors are adopted narrowly as efficiency tools, they could further weaken already fragile forms of shared intellectual life. However, they could also create opportunities for institutions to reinvest more deliberately in the human dimensions of higher education: community, collaboration, mentorship, identity formation, and shared intellectual experience.

Recent discussions in higher education have primarily focused on GenAI as a tool for creativity, feedback, advising, instructional efficiency, and the ethical and institutional questions such tools raise.^Footnote1 At the same time, emerging scholarship on agentic AI increasingly describes the building blocks of more autonomous systems, including planning, tool use, persistent memory, and adaptive decision-making.^Footnote2 Yet these conversations often remain fragmented. They describe capabilities, applications, or perceptions but stop short of considering how those might converge into a sustained instructional role. The Agentic Professor occupies that space, raising questions not about how AI supports individual tasks but how higher education might change when AI becomes a persistent intellectual presence throughout a student's educational experience.

Providing Individualized Instruction at Scale

Envisioning the Future, a first-year course taught by one of the authors at the University of Rhode Island until his retirement in 2022, examined emerging technologies and their potential societal implications. It introduced students to technologies likely to shape their adult lives, including AI, robotics, genetic engineering, quantum computing, blockchain, synthetic biology, lab-grown organs, and novel food systems. Its goal was to help students develop not technical mastery but conceptual fluency: the ability to reason about emerging technologies, their broader implications, and their trajectories.

Teaching the course was exhilarating—and unexpectedly difficult. It exposed a structural tension present throughout undergraduate education: the gap between what high-quality, individualized instruction demands and what human instructors can realistically provide at scale. Initially, guest lecturers were invited to distill foundational discipline-specific material. However, these faculty members often gravitated toward the aspects of their disciplines they personally found most compelling rather than the conceptual information students needed. Relying on a single instructor solved that problem but created another: a single instructor now had to cover distinct technological areas, each requiring students to learn new concepts, language, and frames of reference, even as underlying fields advanced faster than any one person could confidently track. Student diversity compounded these challenges. First-year cohorts arrived with widely varying levels of quantitative preparation, reading and writing ability, conceptual reasoning skills, exposure to science and technology, and motivation.

This tension between the ideal of individualized instruction and the constraints of scale is not new. In 1984, educational psychologist Benjamin Bloom described the 2 sigma problem: students who received one-on-one tutoring performed two standard deviations above those in conventional classrooms, meaning the average tutored student outperformed 98 percent of traditionally taught peers.^Footnote3 Yet Bloom also recognized that individual tutoring was "too costly for most societies to bear on a large scale." For more than forty years, addressing that gap has been a goal of educational technology, a goal that has not yet been achieved. The Agentic Professor as conceived herein would address this gap—and more. These constraints motivate the question of how higher education changes when sustained, individualized guidance becomes scalable through an agentic, long-horizon instructional partner.

Understanding the Pedagogical Nature of the Agentic GenAI Professor

Recent scholarship on agentic AI in education provides an important foundation for this discussion. Researchers have described agentic AI functions such as autonomous goal pursuit, adaptive pedagogical planning, multimodal interaction, learning analytics, and persistent learner modeling.^Footnote4 Early forms of these features already exist, albeit in limited and fragmented implementations. The Agentic Professor extends this trajectory not as a description of current systems but as a plausible configuration that combines these elements into a sustained instructional role.

It is tempting to imagine the Agentic Professor as a more capable chatbot or tutor, but such analogies miss the central shift. Most AI tools currently deployed in education are reactive, responding to prompts, questions, or submissions. Agentic systems, meanwhile, can initiate inquiry, propose learning pathways, identify misconceptions, and dynamically adapt pedagogical strategies. This is not an incremental improvement; it is a categorical change in how instruction unfolds.

Just as importantly, instruction is anchored in student-generated questions, giving students a sense of agency while the Agentic Professor retains pedagogical control.

The power of this interaction is not mysterious. It mirrors the best Socratic teaching methods used by skilled human instructors. The Agentic Professor offers this level of engagement consistently, patiently, and at scale, tailored to each student's interests and reasoning patterns.

And because the Agentic Professor retains a detailed memory of what engaged a particular student, the episode becomes a reusable reference point: the skipping-stone/baseball comparison can be productively resurfaced weeks, months, or even a couple of years later in a new setting, strengthening continuity and transfer. Another student, presented with the same initial question from the Agentic Professor, may anchor on an entirely different example that is equally "sticky" for them—giving the Agentic Professor a personalized thread it can carry across topics and time.

Although illustrated here in physics, the same structure could be applied in literature, art, history, and other disciplines. The Agentic Professor could begin with the student's point of affinity and use it to anchor learning, helping students engage with unfamiliar concepts through interests and experiences that are personally meaningful.

Interactions of this kind are not entirely hypothetical. Current GenAI systems can already reproduce elements of this adaptive, probing exchange, although without the sustained agency, memory, and continuity envisioned for the Agentic Professor. Readers can experience some of these capabilities through the exercise in the experiment described below.

What distinguishes the Agentic Professor envisioned here from such present-day interactions is not any single function but the combination of capabilities sustained over time: agency, longitudinal memory, pedagogical plurality, cross-disciplinary synthesis, and continuous refinement. Its significance lies not simply in providing individualized support but in creating continuity across a student's educational experience, connecting ideas, skills, and feedback across courses, disciplines, and years of study. A course like Envisioning the Future can gesture at that synthesis within a single semester; no course structure can sustain it across four years.

The Agentic Professor's capacity for sustained, individualized guidance reconfigures the relationship between teaching and assessment. Teaching and evaluation could merge into a single, ongoing process. As students reason aloud, revise explanations, and justify choices, assessment becomes an organic byproduct of learning rather than a separate, episodic event. This integration, however, would require substantial changes to accreditation frameworks, grading policies, and transcript formats.

The Agentic Professor is introduced here as a one-on-one instructional partner, but there are other natural extensions. One possibility is a second Agentic Professor that observes the exchange and provides a "second opinion," surfacing misconceptions, blind spots, or missed opportunities without engaging the student directly—an automated form of quality assurance. Another possibility is a single Agentic Professor who interacts simultaneously with a small group of students, potentially reducing the social inhibition associated with faculty authority and encouraging students to test ideas and expose uncertainty. Such an approach could preserve, and perhaps enhance, group interactions that are central to many learning experiences.

These examples illustrate that the Agentic Professor does not occupy a fixed role but instead functions as a flexible framework that can be adapted to different pedagogical contexts.

Rethinking Assessment, Governance, and Institutional Responsibility

Looking ahead requires some simplifying assumptions. References to a four-year degree in 2036 should be understood as a baseline for discussion rather than a prediction. Undergraduate education may evolve significantly over the next decade, and students entering college may differ markedly from today's cohorts, particularly if many have already learned with agentic teachers in K–12 settings. Those changes could reshape student preparation, expectations, and instructional needs, which means the institutional implications of Agentic Professors cannot be assessed by assuming today's students and degree structures persist unchanged.

The question that naturally follows is how many instructional functions such a system might assume. One plausible path is that human faculty and academic governance remain accountable for the ethical, pedagogical, and intellectual architecture of academic programs, while the Agentic Professor executes aspects of that architecture with a depth and continuity that would be difficult for a human instructor to sustain for every student. Other options are possible, of course, raising corresponding questions about legitimacy and oversight.

More consequential than what these systems can do is the question of how higher education institutions define responsibility, accountability, and authority. Agentic AI may become deeply integrated into teaching and learning, but the extent to which colleges and universities legitimize those activities as "instruction" and where they draw the line between delegation and governance will remain institutional choices. If evaluation becomes continuous rather than episodic, for example, institutions will need to rethink assessment artifacts, such as grading policies, credit-hour conventions, and transcript formats.

Those choices are likely to be shaped not only by technical capabilities but by accreditation requirements, liability concerns, faculty governance, and broader social, economic, and political pressures. When technologies challenge perceptions of human uniqueness, agency, or authorship, resistance can become a decisive factor, particularly in societies that place strong value on individual autonomy.^Footnote5

Even when technological capabilities advance and cultural forces favor change, higher education rarely changes in the clean, linear way that disruption narratives suggest. In the early 2010s, for example, many predicted that massive open online courses (MOOCs) would disrupt higher education entirely. They did not. MOOCs scaled content but not connection. Completion rates were low, feedback was sparse, and learning was shallow for most participants.^Footnote6

In hindsight, MOOCs failed not because the technology was weak but because the model ignored how humans learn. As Reich and Ruipérez-Valiente observed, "New education technologies are rarely disruptive but instead are domesticated by existing cultures and systems."^Footnote7The Agentic Professor represents the inverse proposition: connection that scales and personalization without loss of depth. Yet the lesson remains: technological plausibility does not ensure institutional adoption.

Cultural norms, faculty identity, accreditation frameworks, political pressures, and deeply held beliefs about what education should be will shape which futures materialize. The Agentic Professor will encounter resistance not because it fails pedagogically but because it challenges symbolic and social boundaries that institutions are reluctant to cross, such as faculty members' sense of professional value, identity, and authority.

Preparing for a Future That Will Not Wait

The Agentic Professor represents one possible response to a deeper reality: higher education institutions are encountering a technology that may challenge longstanding, foundational assumptions about how instruction is organized, how learning unfolds over time, and the role of faculty. Yet the Agentic Professor does not replace the broader functions of colleges and universities. Instruction may transform, but higher education institutions remain social institutions where identity is formed, communities are built, and values are transmitted through human interaction.

AI capabilities will continue to advance rapidly, and many traditional instructional functions overlap directly with those capabilities: explanation, feedback, evaluation, advising, research support, and even motivational coaching. The question is not what AI can do but what higher education institutions choose to preserve, redesign, and delegate—and why.

Engaging seriously with ideas like the Agentic Professor does not commit institutions to take a particular path. But refusing to engage guarantees that decisions will be made elsewhere—by students, by employers, by markets, and by technologies that evolve faster than tradition can accommodate. Resistance will be real and consequential, shaped by faculty identity, parental trust, political narratives, and accreditation requirements. The pace of AI development ensures that delay does not preserve the status quo; it merely postpones reckoning.

Multiple futures are possible, and none is inevitable. The conversation is already underway; what remains is whether colleges and universities shape it deliberately—or inherit it by default.

Acknowledgment

The authors thank P. Matthieu Cornillon for feedback that sharpened the framing and strengthened the institutional analysis.

Notes

1. See, for example, Brian Basgen, "AI as a Thought Partner in Higher Education," EDUCAUSE Review, April 9, 2025; Danny Y.T. Liu and Simon Bates, Generative AI in Higher Education: A Framework for Action and Future Innovation (Association of Pacific Rim Universities, January 14, 2025); Ted Brodheim, "Shaping the Future of Learning: AI in Higher Education," EDUCAUSE Review, May 12, 2025; Yueqiao Jin et al., "Generative AI in Higher Education: A Global Perspective of Institutional Adoption Policies and Guidelines,"Computers and Education: Artificial Intelligence 8 (June 2025): 100348. Jump back to footnote 1 in the text.↩
2. See, for example, Georgios Kostopoulos et al., "Agentic AI in Education: State of the Art and Future Directions," IEEE Access 13 (2025); Samar Aad and Mariann Hardey, "Generative AI: Hopes, Controversies and the Future of Faculty Roles in Education," Quality Assurance in Education 33, no. 2 (2025): 267–282.Jump back to footnote 2 in the text. ↩
3. Benjamin S. Bloom, "The 2 Sigma Problem: The Search for Methods of Group Instruction as Effective as One-to-One Tutoring,"Educational Researcher 13, no. 6 (June/July 1984): 4-16; Bloom's "2 sigma problem" is developed primarily from K–12 tutoring and mastery-learning studies. It is cited here conceptually rather than literally, not as a direct claim about undergraduate effect sizes but as a durable framing of the same constraint highlighted in this case study: individualized, high-touch instruction is difficult to provide at scale. The relevance of Bloom's framing to postsecondary contexts is discussed in Robert S. Feldman, ed., The First Year of College: Research, Theory, and Practice on Improving the Student Experience and Increasing Retention (Cambridge University Press, 2017), chap. 8. Jump back to footnote 3 in the text.↩
4. Deepak Bhaskar Acharya et al., "Agentic AI: Autonomous Intelligence for Complex Goals—A Comprehensive Survey," IEEE Access 13 (2025): 18912-18936; Soodeh Hosseini et al., "The Role of Agentic AI in Shaping a Smart Future: A Systematic Review,"Array 26 (2025): 100399; Abhishek Raidas and Ravi Bhandari, "Agentic AI in Education: Redefining Learning for the Digital Era," in Artificial Intelligence in Education: Transforming Learning for the Future (A2Z EduLearningHub, 2025), 89–98.Jump back to footnote 4 in the text. ↩
5. Brian Kennedy et al., "How Americans View AI and Its Impact on People and Society," Pew Research Center, September 17, 2025. Jump back to footnote 5 in the text.↩
6. Justin Reich and José A. Ruipérez-Valiente, "The MOOC Pivot," Science 363, no. 6423 (January 11, 2019): 130–131. Jump back to footnote 6 in the text.↩
7. Ibid. Jump back to footnote 7 in the text.↩

---

Peter Cornillon is Emeritus Professor of Oceanography at the University of Rhode Island.

J. Xavier Prochaska is Professor at UC Santa Cruz.