Key points:
- Higher education cannot bolt AI onto a curriculum built for a different era and call it innovation
- Colleges are adopting AI faster than they can govern it
- Dissecting higher ed’s complex–yet promising–relationship with AI
- For more news on higher education and AI, visit eCN’s AI in Education hub
Higher education is having a familiar conversation in an unfamiliar moment. We are debating whether students “should” use AI, whether it is “ethical,” whether it is “cheating,” whether we can “ban” it, whether we can “detect” it, whether it will “go away.” This is what happens when an institution confuses discomfort with principle.
AI is not a fad. It is not a tool students will abandon when we wag our fingers harder. It is a general capability that is rapidly becoming embedded in the knowledge economy, in professional workflows, and in the daily expectations placed on modern graduates. If our curriculum assumes a pre-AI world, then we are teaching history while pretending it is preparation.
José Antonio Bowen captures the core cognitive dissonance with a line that should sting a little: “What Higher Education calls cheating, Business calls progress.” The point is not that anything goes. The point is that our current moral panic often substitutes for curriculum design. We cling to familiar assignments and then blame students for living in the world as it is, rather than as we wish it were, or it used to be. If we are honest, much of what we call “cheating” is simply students using the most powerful available tool to complete tasks that never demanded much thinking in the first place.
The bluebook retreat is a confession, not a solution
A growing number of faculty are responding to AI by retreating to bluebooks, pencil exams, and proctored in-class writing as the primary proof of learning. Let’s call this what it is: an admission that our assessments are brittle.
When the best solution we can imagine is an archaic technology, we are not defending academic standards. We are reinforcing a quiet but devastating message to students: the education you are receiving is built for the 1700s, not the 21st century.
Bluebooks are not sacred. They are a workaround. They measure performance under constraint, not competence in context. They reward speed, compliance, and recall. They tell students that what matters is what you can produce alone, on command, in a sealed room, with a sharpened pencil and no tools. That is not how modern work happens. It is not how modern knowledge is produced. And it is certainly not how modern problem-solving unfolds. If our only defensible measure of learning is what students can do without any of the tools they will have in their careers, then we are not designing education. We are staging a reenactment.
If AI can answer your question, you might be asking the wrong question
This is the blunt truth faculty need to hear, and I say it with deep respect for the profession: If a prompt can reasonably answer your exam question, your discussion post, or your weekly reflection, the problem is not the student. The problem is the question.
Too many of our assessments still reward recall, formulaic writing, and the performance of comprehension rather than comprehension itself. We assign work that can be completed by pattern matching, then act surprised when a pattern-matching machine does it well. Academic rigor does not mean “harder.” It means “deeper.”
The presence of AI forces us back to first principles:
- What do we actually want students to be able to do?
- What must they understand to do it well?
- What human judgment is essential?
- What types of errors are most dangerous?
- What does excellence look like when a student has a powerful assistant?
When faculty ask, “How do I stop students from using AI?” they are often asking the wrong question. A better question is, “How do I design learning so that AI use reveals thinking rather than replacing it?”
The human-centered approach: Teach students to use AI, not hide it
I am not arguing for AI as an academic shortcut. I am arguing for AI as a human-centered tool inside a curriculum that expects more from students, not less.
Students should be asked complex questions, then taught how to use AI to support the kind of work we claim higher education uniquely develops: analysis, synthesis, critique, creativity, ethical reasoning, and disciplined judgment.
In practice, that means shifting from assignments where the product is the proof, to assignments where process, evidence, and decision-making are visible.
Ask students to:
- Interrogate AI output: Identify claims, check sources, test assumptions, and surface errors.
- Compare prompts and outputs: Show how inputs shape outputs, then justify a final approach.
- Use AI as a collaborator: Generate options, then defend choices with evidence and reasoning.
- Document their workflow: Prompts, revisions, critiques, and the rationale behind edits.
- Apply disciplinary standards: Demonstrate that they know what “good” looks like in the field.
- Reflect on ethics and impact: Privacy, bias, transparency, and accountability.
A student who can produce a polished paragraph is not necessarily educated. A student who can evaluate, improve, and responsibly apply machine-generated work, while demonstrating independent judgment, is.
What we did at San Francisco Bay University
At San Francisco Bay University, we decided that incremental policy updates were not enough. One cannot bolt AI onto a curriculum built for a different era and call it innovation. We redesigned our entire General Education curriculum to infuse AI assignments into every class. Not as an “AI week.” Not as an optional add-on for the faculty who are curious. Not as a workshop series that only the already-enthused attend. Every course includes structured, scaffolded experiences in using AI responsibly and effectively, tied to clear learning outcomes.
Because SFBU is a startup university, we had a unique opportunity that most institutions do not: We could innovate at the institutional level. We could reconceive the curriculum from scratch, and infuse AI wherever appropriate, not wherever convenient or politically viable.
We did not have to retrofit a decades-old general education structure built for a different economy. We did not have to negotiate with legacy systems at every turn. We could build for the world our students are actually entering. I admit that SFBU’s unique Silicon Valley startup culture is a real edge. But it does not excuse everyone else. It simply changes the playbook.
We also purchased an enterprise license of ChatGPT Edu for students, faculty, and staff. That decision was not about novelty. It was about equity and infrastructure. If AI literacy is part of the learning environment, then access cannot depend on who can pay, who has the newest device, or who happens to know which tools exist. We built what students actually need: support that matches the reality of their lives.
Each class has a trained AI socratic tutor available to students 24/7 that speaks 300 different languages and is expert in every critical high-impact pedagogical practice. When a student needs help at 11 p.m. after a shift at work, the answer cannot be, “Come to office hours on Tuesday.” When a multilingual learner is stuck translating a concept, the response cannot be, “Try harder in English.”
We must recognize that AI does not implement itself. Culture does. SFBU has a full-time AI Strategist on staff to help translate technological possibility into curricular reality. And we have mandatory pedagogical training for all full-time and part-time faculty, offered by our Chief Learning Officer. Not optional. Not implied. Required. The most important AI decision a university makes is not which tool it licenses. It is whether faculty are supported and expected to redesign learning around the tool’s existence.
What traditional institutions can do without becoming startups
Most traditional institutions cannot innovate on the institutional level quickly, and anyone who says otherwise has not sat through a faculty committee cycle. However, legacy is not destiny. Even large institutions have leverage points, and they recur with predictable rhythm. Use them.
Here are three high-yield opportunities where real change is possible without pretending the whole university will reinvent itself overnight:
1) Periodic general education revisions
General education is your common core. If it stays pre-AI, your institution stays pre-AI. When you revise GE, do not merely adjust course lists, credit counts, or categorical titles. That’s just old wine in new bottles. Rather, rewrite outcomes around AI fluency, ethical reasoning, human-centered problem-solving, and evidence-based critique, and try to do it in 6-months or less. Embed AI assignments, reflective disclosure, and discipline-specific standards.
Accreditation already demands evidence of student learning, continuous improvement, and alignment with mission and outcomes. AI can be framed honestly inside that language: a modernization of learning outcomes, assessment methods, and student support structures. Use the cycle to build institutional permission for change, and to formalize governance and guardrails. I also remain hopeful that institutional accreditors will eventually modernize and begin requiring AI literacy as a core graduation outcome.
New programs are where institutions quietly innovate because they have to. Bake AI into program design from day one: competencies, assessment plans, internship expectations, tool access, and faculty development. Treat AI literacy like you treat writing across the curriculum: not one course, but a thread. Incremental does not mean timid. It means strategic.
The prescriptive path to becoming AI-first
An institution does not become AI-first by writing a policy memo and hosting a panel. AI-first status is achieved by deciding, at the institutional level, that AI literacy is a core educational responsibility, then building the systems to deliver it.
Here are concrete moves any institution can make now.
1) Stop aiming for AI-proof assessments
Trying to outsmart AI is a losing game. AI detection tools are deeply flawed. Design assessments that make thinking visible: evidence trails, critique and verification, iterative drafts with rationale, oral defenses, authentic performances, and personalized application to real contexts. If students use AI, make them show their work the way we ask mathematicians to show theirs.
2) Make faculty development required and practice-based
Optional training produces optional adoption. Provide model assignments, shared rubrics, and discipline-specific examples. Give faculty time and support to redesign learning. Reward pedagogical innovation with the same seriousness you reward research.
3) Create governance that is real, not symbolic
Define standards for privacy, disclosure, and appropriate use. Provide institutionally supported tools with clear guardrails. Make equity central by providing access, not assuming it.
AI can offer round-the-clock tutoring, multilingual assistance, and structured practice. But the goal is not to replace human relationships. It is to remove avoidable friction so faculty and staff can do the work that only humans can do well: mentoring, motivation, judgment, and care.
The coming shift: AI faculty avatars and the rehumanization of professors
The next pedagogical paradigm is not simply students using AI. It is AI becoming a first-class instructional presence. One of many emerging innovations we are developing thrusts
AI-powered faculty avatars to center stage. These faculty know far more than the most knowledgeable professor. They will be trained to be the best version of the best professor anyone has ever had. They will grade without bias, provide immediate and customized feedback, adapt and customize their approach in real time to the learning preferences of each student, and be available around the clock.
That reality will unsettle the profession, and it should. Not because it eliminates the need for professors, but because it forces a sharper definition of what professors are for, and what they do well. So what then will be the role of the human professor?
It will evolve into doing what faculty do best:
- guiding students through ambiguity
- mentoring identity and purpose
- encouraging perseverance
- counseling through setbacks
- problem-solving in human teams
- modeling intellectual virtues like humility and courage
- teaching students how to disagree well
- helping students translate learning into lives and careers
- giving career advice that is rooted in human understanding
AI will not eliminate the professor. It will expose the difference between information delivery and education. The provocation is simple: Higher education can either lead this shift or be led by it. Keeping our heads in the sand is not a strategy. It is a choice, and one that does not serve students.
Dr. Brad Fuster is the Provost at San Francisco Bay University.