> ## Content Index
> Fetch the complete content index at: https://www.thedelatorrereview.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# What Is a University For in the Age of AI? Education, Employment, and the Value of Connection
- URL: https://www.thedelatorrereview.com/what-is-a-university-for-in-the-age-of-ai-education-employment-and-the-value-of-connection/
- Published: 2026-09-21T21:04:22.000Z
- Updated: 2026-09-21T21:04:22.000Z
- Description: AI can make knowledge cheap, but can it replace what universities truly provide? Drawing on Jason Potts’s The University after AI, this article explores how universities must rethink teaching, assessment, credentials, career preparation, research, and the enduring value of human connection.
- Author: Lydia
- Tags: Artificial Intelligence (AI), Future of Education

In my role as Director of Privacy Initiatives at Santa Clara University School of Law, I spend a lot of time thinking about how we should prepare students for a world being reshaped by AI. I also teach AI Governance, so these questions are not abstract for me: What should students learn? How should we assess what they know? And what should a university provide when knowledge and educational content are increasingly available through AI?

My perspective is also shaped by my own experience as a student. When I returned to school to earn my LL.M. at Santa Clara Law, my goal was clear: I wanted to get a job. I do not think that has fundamentally changed for most students. **Universities serve important intellectual and social purposes, but students also expect their education to help them build a career.**

That value does not come from coursework alone. **The connections a university makes possible—with classmates, professors, alumni, employers, and the broader professional community—are essential.** In my own case, Santa Clara became much more than the place where I earned a degree. It became a professional and intellectual home that has continued to shape my career. In my view, AI may make educational content easier to access, but it cannot easily recreate that community, the trust built within it, or the opportunities that come from belonging to it.

In any event, I am interested in different—and sometimes provocative—perspectives on how universities and legal education should evolve. In *The University after AI*, economist Jason Potts asks a fundamental question: What do universities actually provide, and what happens when AI challenges that value? I may not agree with every part of his analysis or every proposal, but his paper offers a useful framework for thinking about what universities should preserve, what they need to change, and what their role should become in the AI era.

Below is an—admittedly AI-generated—summary of his paper.

### The Thesis of *The University after AI*

In *The University after AI*, Jason Potts, Professor of Economics at Alfaisal University, offers an economic framework for understanding why artificial intelligence poses a fundamentally different challenge to universities than earlier technological innovations. Previous technologies—from the printing press to radio, computers, search engines, and online courses—dramatically reduced the cost of producing and distributing educational content without substantially changing the university. Potts argues that this is because content has never been the university’s core product. Universities principally create value by matching students, scholars, employers, governments, and funders and by credibly verifying knowledge and ability for third parties. AI is different because it disrupts both of those functions (Potts 2026, 2–4).

The paper’s **central thesis is that AI threatens the traditional university not primarily by replacing professors or making instruction cheaper, but by undermining the institutional and economic foundations of its business model.** Potts characterizes the university as a multi-sided platform coordinating at least fifteen constituencies through an elaborate system of cross-subsidies. Its most important asset is its ability to issue a credible “warrant of quality” over education—a product whose quality students, employers, and others cannot directly observe. AI is the first technology capable of weakening both the university’s coordinating role and the credibility of that warrant (Potts 2026, 7–11).

Potts identifies four mechanisms through which this disruption occurs. 

- First, **AI produces what he calls “signal collapse.”** A university credential traditionally functions as a signal because producing the work necessary to obtain it is easier for students with the underlying abilities the credential is supposed to represent. AI lowers the cost of producing essays, analyses, code, and other assessed work for both stronger and weaker students. Even when students comply with institutional rules, the resulting work may therefore reveal less about their independent abilities. The problem is not limited to cheating: AI weakens the relationship between assessed performance and the capability the credential purports to certify (Potts 2026, 7–8, 11–12).
- Second, **AI facilitates disintermediation**. AI systems can increasingly provide educational content, tutoring, research assistance, literature synthesis, and informational support without the university serving as intermediary. This development threatens the student-fee revenue that commonly subsidizes research and other university functions. At the same time, AI reduces researchers’ dependence on institutional infrastructure by substituting for research assistants, database support, computation, and other forms of scholarly assistance. Universities may consequently be disintermediated from both the students who fund the institution and the scholars who produce its research (Potts 2026, 12–13).
- Third, **AI relocates the university’s essential strategic asset**. Administration has historically been important because it operates the systems through which universities coordinate their constituencies and assure quality. AI can automate much of that administration, but it cannot itself provide the institutional warrant on which the credential depends. A credible warrant requires an independent and durable institution that places its reputation at risk when it certifies a person’s capabilities. The university’s indispensable asset therefore shifts from administering the quality-assurance process to serving as a credible verifier (Potts 2026, 13–14).
- Fourth, **AI changes where economic value accrues.** As knowledge transmission and competent execution become cheaper, value moves toward the ability to adapt quickly to new technologies, exercise judgment, verify AI-generated work, and explore questions outside established domains. Universities can remain valuable, Potts argues, if they stop treating access to expert knowledge as the scarce product and instead develop students’ capacity to adapt, judge, verify, and explore (Potts 2026, 14).

Not everything universities provide is equally vulnerable. Potts identifies three important ***“invariants”***: **education remains a process of co-production requiring the student’s participation; AI cannot manufacture the peer relationships and networks created by a carefully composed cohort; and it cannot replace activities requiring embodiment, physical facilities, legal standing, accreditation, licensure, or long-lived institutional trust.** These surviving functions indicate where universities can continue to create distinctive value (Potts 2026, 15).

Based on this analysis, Potts advances **ten proposals for restructuring universities**, organized around three questions: what universities must continue to own, what they should build, and what they should reshape. Together, the proposals envision a university focused less on delivering information and more on providing credible verification, professional preparation, valuable human networks, and forms of research that markets and AI systems are unlikely to produce on their own.

1. **Institutionally separate certification from instruction.** Potts argues that the university’s most important asset in the AI era is not its teaching but its ability to issue a credible “warrant” that a person possesses particular knowledge and abilities. He therefore proposes separating the institution that teaches students from the body that examines and certifies them. Instruction could still be provided by the university, obtained elsewhere, or pursued independently, while students would pay primarily for rigorous assessment and certification. This structure could support examination-only pathways to a degree and would resemble professional certification systems such as the CFA, ACCA, and medical or actuarial boards. To preserve its credibility, the certifying body would need separate governance and funding, protected examiners, and no financial interest in how many students pass (Potts 2026, 16–17).
2. **Restore the “costly signal” of individual ability.** If AI allows students of different ability levels to produce similarly polished essays, analyses, and code, conventional assignments no longer communicate as much about what an individual student can do. Potts considers greater investment in AI detection and punishment an ultimately unwinnable arms race. Instead, he proposes moving assessment into forms in which performance remains meaningfully dependent on the student’s own capabilities. These could include oral examinations, live problem-solving, practical demonstrations, and structured debate or disputation. Such assessments would be more expensive and less scalable, but Potts’s model would use the savings generated by inexpensive AI-assisted instruction to finance more intensive human examination. The purpose would be not simply to prevent cheating, but to test judgment, intuition, taste, and the ability to defend one’s reasoning in real time (Potts 2026, 17–18).
3. **Rebuild the “missing junior loop.”** Entry-level professional work has traditionally allowed recent graduates to learn through relatively routine tasks performed under supervision. AI is beginning to automate precisely those tasks, potentially eliminating the lower rungs through which inexperienced workers acquire professional judgment. At the same time, better screening tools may reduce employers’ incentive to invest in training junior employees. Potts proposes that universities fill this gap by extending education into the early stages of professional development. They might arrange supervised post-degree placements, establish structured apprenticeship years, or use realistic simulations to develop judgment in fields where traditional junior work is disappearing. Universities would therefore take on some of the training function previously performed by employers, although doing so would require new ways to finance that training and share in the value it creates (Potts 2026, 18–19).
4. **Create fast-response accreditation.** If economic value increasingly flows to people who can adapt quickly to emerging technologies, universities cannot continue taking several years to approve a qualification in a rapidly developing field. Potts proposes standing units with delegated authority to create, certify, and retire short qualifications within weeks rather than years. These programs could address technologies such as AI agents, drones, or cryptocurrency while the relevant skills remain scarce and valuable. The university’s distinctive contribution would not be producing course content, which AI can help generate quickly, but combining disciplinary expertise with the institutional authority to certify that participants have acquired genuine competence. The value of this capability would increase as technological change accelerates (Potts 2026, 19–20).
5. **Treat the cohort as a product.** AI can distribute knowledge, but it cannot easily reproduce the peer relationships and professional networks created by bringing a carefully selected group of people together. Potts therefore proposes that universities deliberately design and market the cohort rather than treating it as an incidental result of admissions. Selection would emphasize not only individual rank but also how students’ experiences and abilities complement one another. Universities would then create structures—such as residential colleges, groups that progress through a program together, sustained collaboration, and dense co-curricular activity—that turn a collection of students into a durable network. This proposal has an important tension: a university that sells broad access at high volume may find it difficult simultaneously to sell the value created by selectivity and scarcity (Potts 2026, 20).
6. **Host scholars rather than necessarily employ them.** As AI reduces scholars’ dependence on research assistants, databases, computing resources, and other institutional infrastructure, some researchers may be able to operate independently. Universities could respond by separating institutional affiliation from full-time employment. They might offer different levels of affiliation that provide access to grant administration, ethics and compliance systems, facilities, colleagues, or the right to publish under the institution’s name. Affiliation would remain subject to quality review because the university would be lending the researcher its reputation. Potts also considers a more extensive restructuring in which teaching, supervision, and other measurable work could be contracted separately, while researchers participate in contribution-based pools that distribute funding and returns according to their work. This would transform some academic employment costs into paid affiliation services while allowing universities to maintain relationships with scholars whom they could not employ conventionally (Potts 2026, 20–22).
7. **Move research to the new frontier.** Potts argues that AI may increase individual researchers’ productivity while narrowing the overall range of inquiry by directing scholars toward fields with abundant data, established methods, and readily measurable results. Universities should therefore concentrate on research that private incentives and AI-assisted scholarship are likely to undersupply. This includes low-data fields, unfashionable questions, negative results, long-term projects, methods that cannot be reduced to prediction, and inquiries into what he calls the “unknown unknowns.” As literature reviews, statistical analysis, and other established research tasks become cheaper, the university’s comparative advantage should move further toward genuinely exploratory work at the edges of existing knowledge (Potts 2026, 22).
8. **Shed sides of the university platform.** Universities coordinate many constituencies and activities, often using revenue from one group to subsidize services provided to another. When revenue declines, the usual response is to seek incremental efficiencies across the institution. Potts proposes a more deliberate form of contraction: universities should determine which constituencies and functions they can serve better than available alternatives and withdraw from those they cannot. Activities might be sold, spun out, or transferred to shared providers. Universities should retain functions that depend on legal authority, physical presence, durable trust, or the institutional warrant, while reconsidering those based primarily on transferring information. Because the university is a system of interdependent cross-subsidies, however, eliminating one function would require restructuring relationships with the remaining constituencies—not merely cutting a department or reducing its budget (Potts 2026, 22–23).
9. **Make the university’s claim on graduate outcomes explicit.** Universities already benefit when their graduates succeed: strong outcomes enhance institutional reputation, recruitment, and alumni giving. Potts proposes converting that indirect relationship into a more explicit financial arrangement. Possible mechanisms include income-share agreements, claims tied to the earnings advantage associated with the degree, cohort-level contribution systems, or a right to invest in or make offers to graduates. This might give universities both the resources and the incentive to support students beyond graduation, including through the extended professional training contemplated by Proposal 3\. Potts acknowledges that this is among his more speculative proposals. It would present serious concerns involving adverse selection, enforceability, liquidity, and the propriety of universities holding financial interests in their graduates. Importantly, any such interest would have to belong to the instructional or training arm—not the independent body certifying the graduate’s abilities (Potts 2026, 23–24).
10. **Treat governance as part of producing a credible warrant.** Potts’s final proposal connects institutional governance directly to the university’s certification function. If the university’s essential product is a trusted warrant of quality, safeguards that make its academic judgments credible should not be viewed merely as inefficient traditions or administrative costs. Peer judgment, academic freedom, protections for examiners, independence from the people being assessed, and the institution’s ability to endure beyond particular leaders are part of the process through which trust is produced. Reforms that make examiners dependent on student satisfaction, pass rates, managerial preferences, or financial outcomes could weaken the university’s most valuable remaining asset. Potts does not argue that every university function must be governed collegially; rather, those protections must attach specifically to the independent certifying body (Potts 2026, 24).

Taken together, these proposals do not describe the traditional university with AI added to its existing operations. They **envision a more fundamental reallocation of institutional resources: inexpensive and potentially decentralized instruction paired with expensive human assessment; shorter and faster credentials; greater responsibility for early-career training; intentionally designed professional communities; more flexible relationships with scholars; and research directed toward questions that AI and market incentives are least likely to pursue.** Potts recognizes that such changes would be difficult within existing universities because altering one part of a multi-sided institution affects all the others. For that reason, he believes these models are more likely to emerge in new universities, campuses, colleges, or separately governed units than through incremental reform of existing institutional structures (Potts 2026, 15, 25).

The paper ultimately rejects the assumption that universities can absorb AI as merely another productivity tool while preserving their existing institutional form. Potts concludes that the most viable reforms are likely to occur through new universities, colleges, campuses, or separately governed units designed around credible examination, rapid adaptation, and exploration—not through incremental reform of incumbent institutions. In his account, **the university can survive AI, but only by recognizing that its enduring product is not instruction or content. It is the credible verification of human knowledge, judgment, and capability** (Potts 2026, 25).

### Source

Jason Potts, *The University after AI*, Version 0.1, September 9, 2026, available through [SSRN](https://papers.ssrn.com/sol3/papers.cfm?abstract%5Fid=7453440&ref=thedelatorrereview.com).