AI in Higher Education Newsletter
May 09, 2026 · Vol. 16
A weekly brief for the Management Department, Isenberg School of Management, UMass Amherst. By Matthew D. Langenkamp / 雷邁德, with research assistance from Thea 🪻✨.
← Back to archive · ← Back to Dictionary
Overview
Top story this week is an interesting mainstream article in a trusted publication. The New Yorker published a long-form piece asking whether AI breaks the university’s credentialing function, naming consulting, finance, and management by name. The Class of 2026 walked into the worst job market for new graduates in four years. California State University’s $17M OpenAI deal — signed without student or faculty consultation — became the cleaner, larger-scale sibling of the ASU Atomic story we covered last week. A Cornell/CMU study inverted the usual equity narrative on AI in admissions. And Canvas — the LMS roughly four in ten North American institutions run on, including ours — was knocked offline by a ransomware attack during finals week.
Five items below. Operator’s note at the close: this issue is being written on Saturday morning rather than Friday, the consequence of a misfired automation that has now been redesigned. Vol. 17 will return to the regular Friday cadence.
1. The Credentialing Question Goes Mainstream: The New Yorker Asks Whether AI Makes College Obsolete
Jay Caspian Kang opened a multi-week New Yorker series on May 5 with the question of whether AI — on top of federal funding cuts, rampant cheating, and accelerating job displacement — finally breaks the university’s credentialing function. The data points he marshals are familiar to anyone who has been reading the trade press, but the venue is new: more than one in four college students say their tuition was not a good investment; more than 40% of college graduates between 22 and 27 hold jobs that don’t require a degree; seven in ten Americans believe higher education is “going in the wrong direction.” Kang quotes a specific list of professions he sees as exposed: “whether they’re in consulting, insurance, finance, management, or the sciences.” That list will look familiar. (The New Yorker, May 5, 2026)
Kang is, by his own framing, cautiously skeptical that the system will actually collapse — universities have outlasted pandemics, wars, and MOOCs — but he is unmistakably treating the current pressures as extraordinary. The significance of the piece is less in its conclusions than in its venue. The New Yorker is read by donors, parents, board members, and policymakers. When the credentialing question moves from Inside Higher Ed to The New Yorker, the conversation changes: the question is no longer “are we keeping up with the trade-press chatter,” it is “what will we say when a board member emails us this article.”
For this department, the relevant move is to be ready with a substantive answer to a now-mainstream question. The four professional categories Kang names — consulting, finance, management, sciences — are the categories Isenberg’s degree leads into. What does an Isenberg graduate do that AI cannot? is no longer a faculty-lounge question; it is a question we will be asked. AACSB Standard 4.1 (effective July 1) gives accreditation reviewers the hook: curriculum must “reflect current and emerging business theories, technologies, and practices.” A defensible answer is not “we use AI in some classes.” A defensible answer is a coherent account of what durable value the degree adds in an AI economy.
2. Cal State’s $17 Million OpenAI Deal — and What “Shared Governance” Now Means
CalMatters reported on May 1 that California State University — the largest public university system in the U.S., with over 470,000 students — signed a $17 million contract with OpenAI for ChatGPT Edu without consulting either the system-wide student association or faculty governance bodies. The contract is now up for renewal. Katie Karroum, vice president of the Cal State Student Association: “We were not consulted when the contract was signed, and we weren’t even given a heads up.” Some faculty are declining to use the tool on principle. The chancellor’s office cited cost as the deciding factor. (CalMatters, May 1, 2026; pickup through May 6)
This is the same shape as the Arizona State Atomic story we covered last week, scaled up. The pattern is now visible enough to name: major public university systems are procuring AI tools at the institutional level without engaging shared governance, treating AI deployment as a technology-procurement decision rather than an academic-policy one. AAUP’s Spring 2026 issue (Daniel Greene’s piece, “What Does AI Do?”) frames this structurally — Baumol’s cost disease makes AI-as-labor-discipline perpetually attractive to administrators answering to regents, donors, and bond markets. Whether one agrees with the AAUP framing or not, the cost-disease analysis is a legitimate economic argument and would not be out of place in a strategy seminar.
The equity argument cuts both ways. Proponents say equal institutional access to AI tools narrows a gap that already exists — wealthier students at private universities pay for ChatGPT Plus out of pocket. Opponents say the absence of consent and due-process protections in AI-cheating allegations is itself an equity issue. Either reading lands the same operational point: AI procurement decisions made without shared governance create downstream policy decisions that faculty are then asked to enforce without having shaped. AACSB Standard 5 (Faculty Sufficiency and Qualifications) and the broader shared-governance expectation become directly relevant.
For UMass and for Isenberg specifically, the question worth holding is procedural: how would a $17M institutional AI contract surface for faculty review here, and at what point in the procurement cycle? It is a question better asked before the contract than after.
3. The Class of 2026 Walks Into the Worst Job Market for New Graduates in Four Years
The unemployment rate for recent college graduates has reached a four-year high, according to a Scripps News analysis published on May 7. Job postings targeted at new graduates are declining; AI displacement of entry-level roles is cited as a compounding factor. The story is consistent with the New York Times’s late-April reporting that recent graduates are lowering expectations, accepting roles unrelated to their degrees, and deferring graduate school. (Scripps News, May 7, 2026)
This story is landing in the same week most of our students are receiving degrees. Business graduates entering consulting, finance, and management roles — the historically strongest recruiter categories for Isenberg — are walking into a market that is tighter and more AI-disrupted than any cohort we have placed in the past four years. The “value of the degree” question is not abstract this spring; it is in the room for every advising conversation and recruiting presentation, and it will follow our students into their first six-month performance reviews.
For colleagues teaching capstone or career-adjacent courses, the relevant adjustment may be in how we frame the value proposition we are sending students out with. The degree is not a credential that exempts a graduate from an AI-disrupted market; it is a foundation that — done right — equips them to compete in one. That distinction is worth being explicit about, both with students and with the recruiters and parents asking the question. AACSB’s 2026 Standards include employment outcome data as part of program effectiveness evidence; structural deterioration in the graduate job market is now part of the explanatory environment in which those numbers will be read.
4. The Equity Story on AI in Admissions Runs Opposite to What Most People Assume
A Cornell / Carnegie Mellon study published this week, analyzing tens of thousands of admissions essays at an unnamed selective institution before and after the launch of generative AI tools, found that lower-income applicants — proxied by those receiving fee waivers — were more likely to use AI in their college essays than higher-income applicants. Students who were ultimately rejected were also more likely to have used AI. Lead researcher Jinsook Lee (Cornell) reads the finding as consistent with a simple hypothesis: lower-income students, lacking access to private college counselors and tutors, turn to AI as a substitute resource. The researchers also found a measurable increase in homogeneous language across the essay corpus after 2022 — AI is flattening the stylistic diversity of admissions pools. (Inside Higher Ed, May 8, 2026)
This inverts the usual narrative. The conventional concern is that wealthy students will gain an AI advantage. The actual finding is the opposite: lower-income students use AI more, but it doesn’t help them — they are rejected at higher rates — and the homogeneity in the writing may actually mark them out to admissions readers. The pedagogical implication for our courses is uncomfortable but worth sitting with: blanket AI-prohibition policies on written work may function as a proxy that disadvantages exactly the students already disadvantaged in access to traditional academic support.
The honest move is to think harder about what we are assessing when we assess written work. If the goal is to evaluate whether a student understands the material, AI-assisted production is a problem only insofar as it lets them produce competent text without understanding. The verification gap — can the student demonstrate understanding when asked? — is the design question. Take-home cases without an in-class follow-up conversation no longer test what they used to test. AACSB’s assurance-of-learning framework, which depends substantially on written artifacts as SLO evidence, faces a legitimate validity question in this environment that does not have an easy answer.
5. Canvas Goes Down: A Reminder That AI Runs On Top of Infrastructure
The criminal extortion group ShinyHunters breached Instructure (the company that owns Canvas) the week of May 5, issuing a “pay or leak” ransom demand and forcing Canvas — used by 41% of North American higher education institutions, including ours — temporarily offline. The University of Illinois Urbana-Champaign postponed all final exams scheduled Friday through Sunday. Baylor and Arizona State were affected. Multiple provosts wrote to students with the telling phrase: “the postponement affects all classes — even those that don’t use Canvas.” Service has since been restored; the legal and reputational aftermath is ongoing. (Inside Higher Ed, May 5 and May 8, 2026)
This is not an AI story directly. It is an infrastructure story, and it matters because AI tools in higher education increasingly run on top of the same digital infrastructure that just went down. If Canvas were also running AI-based tutoring, adaptive assessment, or proctoring at the time of the breach, the blast radius would have been substantially larger. The risk architecture of “AI in the classroom” includes the underlying LMS-and-vendor risk that this incident has now made visible to every CIO and provost in the country.
For our department, the operational implication is small but real: assessment continuity planning needs to assume the LMS can be unavailable for days, not hours. The strategic implication is larger: as we move more pedagogically important work onto LMS-delivered AI tools — adaptive tutoring, AI proctoring, grading copilots — the concentration risk grows. Schools that have moved significant assessment weight onto LMS-delivered tools are now operating with concentration risk that this incident has surfaced. It is worth a conversation, before the next breach, about which kinds of student work should not depend on a single vendor’s uptime.
Looking Ahead
Watch for: the next New Yorker installment in Kang’s series — given the venue’s reach to lawmakers, donors, and parents, any piece that names business schools specifically warrants a fast institutional response. Ethan Mollick speaks at the NCCI Annual Insights Symposium May 11–13; his post-conference writing usually carries weight. The 2026 AACSB Business School Data Guide dropped on May 6; the AI-related employment and curriculum figures are worth pulling for benchmarking once we have a chance.
Worth keeping an eye on: Congressional attention on higher-ed vendor cybersecurity standards in the wake of the Canvas breach, and whether state systems begin requiring vendor security audits as a condition of LMS contracts. Both are downstream from a story that finals-week timing made impossible to ignore.
Prepared with assistance from Thea, my AI assistant, who scans the week’s higher-education-and-AI press and prepares a research scaffold each Saturday morning. Final selection, framing, and judgement remain mine. Corrections and reading suggestions: please reply.
— MDL / 雷邁德