Ask The PhD Community

HigherEd AI Daily

September 23 – Provosts Under Pressure, AI Costs in Free Fall

Wednesday, September 23, 2026

Today's edition covers the fractured state of AI governance across campuses, a historic model price competition that could reshape institutional AI budgets and student access, and a growing body of research suggesting that problem-formulation is now the skill higher education needs to prioritize most.

Inside Higher Ed — Governance

2026 CAO Survey: Provosts Navigate Fractured AI Use Amid Funding Pressures

Inside Higher Ed's 2026 Annual Survey of Chief Academic Officers reveals a divided higher education landscape in which academic leaders are largely navigating artificial intelligence on their own, relying on personal AI tools in the absence of unified institutional frameworks. The survey draws responses from hundreds of CAOs across institution types and mission profiles and describes a climate shaped equally by accelerating AI capabilities and mounting political and financial disruption.

Many provosts report making individual decisions about which AI tools to use and how, without coordinated guidance from governing boards, technology offices, or faculty senates. This decentralized approach to AI adoption creates systemic risks around consistency, equity, and accountability; institutions where AI use is concentrated among well-resourced faculty or departments risk widening internal gaps in teaching quality and research productivity.

The broader political environment adds further strain. Federal funding uncertainty and ongoing pressure to demonstrate student outcomes place academic leaders in a difficult position: they must demonstrate innovation while managing tightening budgets, often without the staff infrastructure to support either goal effectively.

Why it matters for campuses

When a national survey of academic leaders describes AI adoption as fractured and largely personal, that is a governance signal, not just a technology concern. Faculty senates, board committees on academic affairs, and offices of academic technology should treat these findings as a prompt to accelerate institutional policy development; the longer informal adoption outpaces formal governance, the harder it becomes to correct inequities and build the accountability structures students and faculty deserve.

Read More

(Subscription may be required)

The Neuron Daily — Tools

GPT-6 and Opus 5.5 Launch a Price War That Could Reshape Campus AI Adoption

Within a 90-minute window on Tuesday morning, Anthropic and OpenAI each released new flagship model families at significantly lower price points than their predecessors. Anthropic launched Claude Opus 5.5 at $4 per million input tokens and $20 per million output tokens, a rate Anthropic says represents roughly a 40 percent cost reduction for typical workloads due to improved token efficiency. OpenAI responded with GPT-6 Sol at $2 per million input tokens and $10 per million output tokens, and GPT-6 Luna at $0.10 per million input tokens and $0.50 per million output tokens; Luna's pricing places high-volume AI use within reach of budgets that could not previously support it.

Independent comparisons published throughout Tuesday found the two model families performing differently depending on task type. Opus 5.5 showed advantages on multi-step reasoning in several tests, while Sol completed coding and agent tasks faster and at meaningfully lower cost. Luna's near-commodity pricing makes it viable for applications where volume matters more than peak reasoning quality, such as large-scale document summarization, classification, and structured data extraction.

Both providers are positioning these model families as substrates for agentic workflows and institutional AI deployments. The competitive dynamic is unlikely to be temporary; pricing will almost certainly continue to fall as model efficiency improves and the market for enterprise AI expands.

Why it matters for campuses

Dramatic cost reductions carry practical implications for institutional AI planning in at least three areas: per-seat subscription pricing for enterprise AI products may face downward pressure as raw token costs fall; institutions building API-based tools for research workflows or student-facing services will find the unit economics substantially more favorable than one year ago; and the access gap between well-resourced and under-resourced students may narrow if campuses adopt lower-cost model families for student tools. Technology officers and academic computing teams should factor continued price compression into multi-year planning assumptions rather than treating today's pricing as stable.

Read More

Inside Higher Ed — Policy

AI-Edited Stanford Ad Violated University Policy, Raising Enforcement Questions Across Campuses

Stanford University confirmed this week that an AI-edited advertisement violated the institution's AI use policy, according to a university official cited in Inside Higher Ed. The episode is notable not because AI policy violations at universities are unusual but because this one was acknowledged publicly; that level of transparency remains the exception rather than the norm in higher education.

Universities have accelerated the production of AI governance frameworks over the past two years, but awareness of those policies among administrative and communications staff is inconsistent. Most institutional AI policies were written with research integrity, academic honesty, and student work in mind; the application of those policies to institutional advertising, marketing communications, and public-facing content is frequently less explicit and less well understood among the staff most likely to use AI tools in those contexts.

The Stanford case raises practical questions about where AI use policies apply, how violations are detected, and what accountability mechanisms exist. Without training, workflow checkpoints, and clear escalation paths, policies that exist on paper may not prevent incidents in practice.

Why it matters for campuses

Offices of marketing, communications, university relations, and advancement are among the most active adopters of generative AI tools at many institutions, yet they are often the least connected to academic AI governance structures. Campus AI policy leads should conduct a targeted review of how policies are communicated and enforced in these offices, and determine whether existing frameworks are sufficiently explicit about AI use in externally published institutional content.

Read More

(Subscription may be required)

The Neuron Daily — Research

As AI Commoditizes Answers, Teaching Students to Find Better Questions Becomes the Priority

A working paper by economists Felix Feng, Brett Green, Curtis Taylor, and Mark Westerfield, highlighted this week, argues that as artificial intelligence makes problem-solving increasingly fast and inexpensive, the relative economic value of problem-finding rises. In a world where competent answers are abundant, the scarce and therefore most rewarded capacity becomes the ability to identify which questions are genuinely worth asking. The argument carries direct implications for curriculum design across disciplines.

A complementary analysis by labor economist Harrison Satcher adds texture to the picture. Satcher argues that current labor forecasting models systematically underestimate AI-driven workforce change because the jobs being created do not yet have established names or occupational categories. The same logic applies to curriculum design: educators cannot train students for roles that do not exist as defined categories, but they can build the habits of mind that allow students to navigate ambiguity, identify meaningful gaps in knowledge, and evaluate AI outputs critically rather than passively.

Taken together, these perspectives reinforce the enduring value of inquiry-based pedagogies. Seminars, capstone projects, undergraduate research experiences, and problem-based curricula may prove more durable investments than any curriculum module focused narrowly on AI tool proficiency, precisely because they develop capacities that AI systems are not positioned to replicate.

Why it matters for campuses

Curriculum committees and provosts weighing AI integration strategies should distinguish between AI literacy as tool training and AI literacy as intellectual preparation. The evidence points toward higher returns from the latter; investing in students' capacity to originate and rigorously evaluate ideas may be a more defensible and durable competitive advantage for graduates than familiarity with any particular AI system.

Read More

(Subscription may be required)

Tool of the Day

LiteParse

LiteParse is an open-source, standalone PDF parsing library built by the LlamaIndex team that runs entirely on your local device with no cloud dependencies or proprietary AI components required. It delivers high-quality spatial text extraction with bounding box support, making it particularly well suited for complex academic PDFs where table structure, column layout, and citation formatting matter. Because it processes files locally, it is appropriate for sensitive documents that should not be routed through third-party cloud services.

For academic researchers and faculty who spend significant time working with research literature, grant documents, accreditation reports, and student submissions, LiteParse offers a lightweight foundation for building AI-assisted reading and analysis workflows without the privacy tradeoffs of cloud-based parsing tools.

Try it: Select five recent journal articles on a topic you are teaching this term, run each through LiteParse to extract clean structured text, then paste the results into a large language model and prompt it to identify recurring methodological assumptions across the five papers. Use the analysis as the foundation for a class discussion on disciplinary conventions and their limits.

Visit LiteParse

Have a great learning day!

Dr. Ali Green

Sources for This Edition

Inside Higher Ed (insidehighered.com)
The Neuron Daily (theneurondaily.com)
TLDR Newsletter (tldrnewsletter.com)

askthephd.com
 | 
askthephd.substack.com
 | 
Unsubscribe

HigherEd AI Daily; Curated by Dr. Ali Green