AI Agents on Campus: Why More Than Half of Higher Ed Staff Are Already Using Their Own AI Tools
A new EDUCAUSE survey of nearly 2,000 higher-ed staff finds 94% already using AI for work -- but only 13% are measuring ROI and 56% have gone around official tools entirely. The classroom debate is missing the bigger story in the back office.
Higher education has spent the last two years arguing about whether AI belongs in the classroom. A new survey suggests the more urgent question is what's already happening outside it -- in the registrar's office, the business office, HR, and IT -- where staff have quietly gone ahead without waiting for an answer.
The 2026 EDUCAUSE report The Impact of AI on Work in Higher Education, published in partnership with AIR, NACUBO, and CUPA-HR, surveyed 1,960 higher-ed staff and faculty between September 29 and October 13, 2025. The headline finding: 94% of respondents said they'd used an AI tool for work in the past six months. This isn't a pilot program anymore. It's ambient.
The Adoption Numbers Look Great. The Governance Numbers Don't
Ninety-two percent of respondents said their institution has some kind of work-related AI strategy, and 81% described their own attitude toward AI as enthusiastic or a mix of caution and enthusiasm. Only 17% said they were cautious or very cautious. Compared to EDUCAUSE's 2024 and 2025 landscape studies, where roughly one in five respondents described institutional leaders as cautious, this reads as a real shift away from wait-and-see.
But three numbers in the same report complicate the good news:
- Just 13% of respondents said their institution is measuring ROI on work-related AI tools.
- Only 54% said they're aware of policies or guidelines governing AI use at work -- meaning close to half aren't.
- 56% said they've used AI tools for work that weren't provided by their institution at all.
Put together: almost everyone is using AI, most institutions have a strategy on paper, but a majority of the people actually doing the work don't know what the rules are, and more than half have gone around official tooling to get something that works better. That's not a policy gap so much as a policy that hasn't caught up to already-normalized behavior.
Where the Time Is Actually Going
When EDUCAUSE asked respondents what they found most promising about AI for their own work, the top answers weren't about teaching or research at all. The three most-selected opportunities were automating repetitive processes, offloading administrative burden and mundane tasks, and analyzing large datasets -- exactly the kind of work that fills a registrar's queue during add/drop week or a financial aid office's inbox in March. From a list of 30 possible work-related AI use cases, 54% of respondents said they'd used AI tools for eight or more distinct types of tasks in the past six months.
That appetite is showing up as habit, not novelty: 73% of respondents who've used AI tools for work do so daily or weekly. Just 11% said they're required to use AI tools for their job, and 64% said they aren't required to and don't expect to be anytime soon -- so this adoption curve is being driven by staff finding the tools useful on their own, not by a mandate from the top.
The Risk List Staff Are Actually Worried About
It isn't blind enthusiasm, either. More than two-thirds of respondents (67%) flagged six or more risks as "urgent." The three risks selected most often were an increase in misinformation, the use of data without consent, and a loss of fundamental skills that come from independent thought. Staff and faculty are using these tools constantly and are simultaneously clear-eyed about what can go wrong when nobody's watching how they're used.
That combination -- high usage, high awareness of risk, low visibility into policy, low measurement of outcomes -- is a specific and fixable shape of problem. It isn't "ban AI" or "let it run." It's that the tools staff already reach for need to be inside a system the institution can actually see, not bolted on around the edges of one.
What This Means for Higher Ed Operations Going Forward
The EDUCAUSE data makes a case that's easy to miss if you're only watching the classroom debate: the more consequential AI rollout in higher ed right now is happening in the back office, one staff member's personal ChatGPT tab at a time. Institutions that want the productivity gains EDUCAUSE's respondents are already reporting -- without inheriting the ROI blind spot and the policy-awareness gap the same survey documents -- need administrative AI use to be something they can actually observe, not something 56% of staff quietly work around.
Workmate
See what an agent team would do for your business.
Keep reading
AI Agents for Museums and Cultural Institutions: What the First UNESCO-ICOM Global Survey Found
UNESCO and ICOM surveyed more than 400 museums across 90 countries and found 57% are already using AI, mostly for administration, while 55% have no internal AI policy at all.
The Agentic AI Productivity Wave Heading for Wealth Management
Deloitte predicts adviser productivity gains of 30% to 100% by 2032 as agentic AI takes over the roughly 70% of an adviser's day spent on behind-the-scenes work, but only 6% of firms use agentic tools today.
AI Agents for HR Teams: Why More Than Half of Companies Still Have None in Place
SHRM's 2026 State of AI in HR report, drawn from more than 1,900 HR professionals, found only 39% of organizations have actually implemented AI in HR — even as 92% of CHROs expect adoption to keep climbing.