The mixed signal colleges are sending
Students are getting two contradictory messages at once about AI, and EDUCAUSE put a precise number on the gap at its 2026 annual conference in Denver. According to the 2026 EDUCAUSE Students and Technology Report, 53 percent of students said their courses had restrictive AI policies, rules that discourage or outright forbid using the tools in coursework. At the same time, those same students are told, correctly and loudly, that employers increasingly want to hire graduates who already know how to work with AI tools competently from day one on the job.
Nicole Muscanell, a senior researcher at EDUCAUSE, used a GovTech interview tied to the conference to name that tension directly and without hedging. Many students are told not to use AI in their courses while employers simultaneously want AI-ready graduates walking in the door, she argued, and that gap between classroom restriction and workplace expectation is now one of the more pointed findings in EDUCAUSE's annual data set. It puts the policy burden squarely back on institutions rather than leaving it to individual faculty judgment calls made course by course.
What baseline AI literacy actually means
Muscanell's central argument is that colleges should stop treating AI literacy as one single, uniform target that applies equally to every student regardless of major or career path. Instead, she argues for a baseline, a floor every graduate should clear before leaving campus, with the required depth above that floor varying considerably by profession and intended career. Her own framing is direct: a base level of knowledge is what everyone needs, but how far past that baseline a student goes depends entirely on the job they are training for.
She offers a concrete contrast between two fields to make the point land. Data scientists may genuinely need advanced capabilities, building, tuning, and evaluating models themselves as part of their daily work. Healthcare workers, by contrast, mainly need to use AI-enabled tools competently and interpret their outputs with appropriate clinical skepticism, never fully delegating judgment to the system. Trying to train every student to the data scientist's depth would waste scarce faculty time on students who will never need it professionally, while under-training the students who genuinely will.
The case against the curriculum overhaul
Perhaps the most useful part of Muscanell's framing for a CTO audience is what she argues against just as firmly as what she argues for: a wholesale overhaul of general education built entirely around AI. That is the instinct a lot of institutions have reached for over the past two years, treating generative AI as transformative enough to justify rebuilding core degree requirements from scratch, and Muscanell cautions against that instinct explicitly in her comments.
Her alternative is considerably less dramatic, and she argues it is also more honest about where the real bottleneck actually sits inside most institutions. Disruption and change are going to keep arriving, in her own words, and institutions that try to lock in a fixed, mandatory AI curriculum right now are effectively optimizing hard for a snapshot of the technology that will likely be outdated before the very first cohort trained under it even graduates.
The real constraint is staffing, not strategy
Muscanell's point that lands hardest for institutional leaders sitting in the audience is that staffing shortages and growing workloads are the actual limiting factor on AI adoption, not any shortage of ideas about what ought to be taught. Faculty need meaningful time to learn new AI tools themselves, hands-on, before they can responsibly integrate those tools into their own teaching practice, and that time has to come from somewhere inside an already stretched weekly schedule that has little slack left in it.
That reframes the AI adoption problem for a CIO or provost from a curriculum design question into a straightforward resourcing question with a real budget line attached. Funding faculty development time, and covering course loads during that development period, is a line item in a budget, not a policy memo that costs nothing to issue. It is also precisely the piece most ambitious AI rollout plans quietly skip in favor of simply announcing the new tool itself.
What the labor market is already signaling
The urgency behind all of this comes from outside the classroom entirely, which is part of why Muscanell leans on it so heavily in her argument. The World Economic Forum's Future of Jobs Report lists AI and big data among the fastest growing skills employers expect to need through 2030, a finding Muscanell cites directly to make her baseline-skills approach feel like alignment with where hiring is genuinely heading, grounded in labor market data rather than academic caution for its own sake.
That external pressure is exactly why the 53 percent restrictive-policy figure matters more than it might appear to at first glance. Institutions whose classroom policies lag meaningfully behind employer expectations are not protecting academic integrity in a vacuum where nothing else changes. They are quietly creating a measurable gap between what their own graduates can demonstrate in an interview and what the job market is already actively screening candidates for during hiring.
The lever a CIO actually has
Read together, the restrictive-policy finding and the staffing warning point to a specific, bounded set of moves for a CIO or vice provost for technology, rather than toward one sweeping institutional mandate handed down from the top. Fund faculty time to genuinely learn the tools before expecting any faculty member to teach confidently with them in front of students. Set one clear, institution-wide baseline AI literacy expectation rather than leaving every individual department to invent its own restrictive default policy in isolation.
Resist the instinct to answer AI anxiety with a brand new mandatory course or a fully rebuilt general education requirement that locks in today's assumptions. The underlying disruption is not finished, so the curriculum investment with the best return right now is the flexible one: a real baseline for everyone, advanced depth only where the job genuinely demands it, and continued faculty time to keep updating both as the technology keeps moving faster than any single curriculum cycle.
