A campus audits itself against the world
Boise State published the results of its AI survey on September 9, and what makes it worth reading beyond the campus newsletter is the benchmark it uses: the Digital Education Council's global AI in Higher Education survey, which drew 45,398 responses, 27,284 from students and 18,114 from faculty, across 35 countries. Boise State's own sample was smaller, 247 student and 117 faculty responses, but the university chose to publish its results specifically in comparison to the global dataset rather than in isolation, which turns a routine institutional survey into a genuine benchmarking exercise.
The headline finding is a split. Boise State scored above peer institutions on faculty adoption, AI-led teaching, and how proactive students and faculty perceive the university's governance approach to be. But it scored below peers on faculty-reported time savings from AI, and both students and faculty expressed more caution than their global counterparts about AI's effect on learning and critical thinking. In plain terms: Boise State appears to be governing AI more carefully than average, but that caution has not yet translated into the productivity gains faculty elsewhere are reporting.
The global numbers behind the comparison
The Digital Education Council dataset that Boise State benchmarked against shows student AI adoption at 88 percent globally and faculty use at 77 percent, up 16 percentage points from the prior year. Those numbers alone would suggest AI has become close to universal in higher education. The more revealing figures sit underneath them: only 29 percent of students believe their instructors can actually guide them on how to use AI well, and 57 percent report inadequate AI guidance specifically around how it affects graded assessments. On the faculty side, just 31 percent say their institution involves them meaningfully in AI policymaking, even though 64 percent have completed some form of AI literacy training.
The US and Canada region stands out as the outlier in the wrong direction. Faculty intent to use AI actually dropped 9 percentage points year over year, from 76 percent to 67 percent, the steepest decline of any region measured. Meanwhile 81 percent of US and Canada students worry AI encourages shallow learning, compared to 66 percent globally, and 38 percent report no AI use is permitted in their assessments at all, against 24 percent globally. As DEC's own leadership put it in describing the findings: "AI has moved into the mainstream of student and faculty life faster than institutions have been able to respond to it. Adoption is now widespread, but coherent practice is not."
Why North America is the outlier
The regional divergence is the part of this data that deserves more attention than it has gotten. It would be easy to assume faculty resistance in the US and Canada reflects unfamiliarity with the technology, but the data points the other way: this region has invested heavily in AI literacy training, and adoption intent is still falling. A more plausible explanation is that faculty who understand AI's limitations well enough are becoming more skeptical of deploying it in high-stakes academic contexts, not less, as their familiarity grows.
That is a meaningful signal for any organization assuming that more training automatically produces more comfortable adoption. It does not, once the people being trained understand enough to see where a tool's failure modes actually sit. Enterprise leaders rolling out AI tools to specialized professional staff, legal, finance, clinical, or engineering teams, should treat rising skepticism among trained users as a data point worth investigating rather than dismissing as change resistance. Sometimes the experts pushing back have found something the rollout plan missed.
The governance-versus-productivity tradeoff
Boise State's specific result, ahead of peers on governance and trust but behind on measurable time savings, maps closely onto a tension enterprise AI programs are already navigating. A cautious, well-governed rollout tends to build institutional trust and reduce the risk of a costly misuse incident, but it also tends to slow the pace at which users discover and adopt time-saving workflows, because guardrails by design limit how freely people experiment. A fast, permissive rollout produces the opposite tradeoff: quicker productivity gains, higher exposure to misuse and governance gaps.
Neither approach is free. What Boise State's data suggests, and what plenty of CIOs have found in their own organizations, is that you cannot currently get both maximum trust and maximum measured productivity gain in the same AI rollout, at least not in year one. The university's own leadership is responding by directing its AI Coordinating Council to focus specifically on "building AI efficiencies for faculty and clearer skill-building pathways for students," an explicit attempt to close the productivity gap without giving up the governance advantage it has already built. That sequencing, govern first and then optimize for productivity once trust exists, is a reasonable template for any enterprise still early in its own AI rollout.
The benchmark worth stealing
The most exportable idea here is not any single statistic but the method: Boise State did not just survey itself, it measured itself against a large external dataset and published where it over- and under-performed. Very few enterprise AI governance programs do this. Most internal AI adoption surveys ask employees how they feel about the company's own rollout, with no external reference point to determine whether the results are good, bad, or simply average for the industry.
CIOs building or refreshing an internal AI governance survey should consider whether a comparable external benchmark exists, or could be constructed from industry survey data, before running another internally-scoped poll. The value in Boise State's results is not that 29 percent of students trust their instructors to guide AI use; it is that the number can be compared against a global baseline and used to identify a specific, addressable gap. Absent that comparison, an internal AI adoption survey mostly just confirms what leadership already suspected, without telling them whether they are ahead of the pack or falling behind it.
