Nearly half of college presidents (48%) now say AI will be the most impactful force in higher education by 2030, ahead of financial pressure and political interference. But only 55% say their own institution is responding to it appropriately, meaning close to half are not confident in their own readiness. That gap between recognizing AI's importance and being able to govern it is not unique to higher education. Deloitte's 2026 enterprise survey found the same pattern industry-wide: agentic AI use is projected to jump from 23% to 74% of organizations within two years, while only 21% report a governance model mature enough to keep pace. Ahead of the IHE Summit's panel on what it means to lead as an "AI-informed" president, the data suggests informed and governed are two different tests, and most presidents have only passed one of them.
Every president now has an opinion on AI. Fewer have an answer for who is accountable when it moves.
What presidents themselves say
Inside Higher Ed's 2026 Survey of College and University Presidents, conducted by Hanover Research across 430 sitting presidents, found that 48 percent now see AI as the single most impactful force facing higher education by 2030, ahead of financial pressure (45 percent) and political interference (43 percent). At the same time, only 55 percent said their own institution is responding to AI appropriately. Read plainly: nearly half of presidents do not feel confident in their own institution's readiness, even as they rank AI as the top strategic force they face.
Is policy catching up?
Some ground is being covered. EDUCAUSE's 2025 AI Landscape Study found that institutions with a formal AI acceptable use policy rose from 23 percent in 2024 to 39 percent in 2025, and 57 percent of respondents now say they view AI as a strategic priority. That is real movement. It also means a majority of institutions still do not have a formal policy in place, a year after AI became the top strategic concern presidents report.
Is this a higher education problem, or a bigger pattern?
Bigger. Deloitte's 2026 State of AI in the Enterprise survey of 3,235 business and IT leaders across 24 countries found that agentic AI use is projected to jump from 23 percent today to 74 percent of organizations within two years, while only 21 percent currently report a governance model mature enough to manage it. The pattern holds across sectors: adoption is accelerating faster than the accountability structures meant to govern it.
The G.U.A.R.D.™ test of an AI-governed president
Being AI-informed means recognizing the force. Being AI-governed means the institution can prove five things, drawn from the G.U.A.R.D. Framework™:
1. Governance. Is there a single named person accountable for AI decisions, documented and communicated? Neither ISO/IEC 42001 nor the NIST AI Risk Management Framework mandates that exactly one person hold this, but a named owner is the clearest way to demonstrate what both are actually asking for. 2. Understanding. Can the institution produce a current inventory of every AI system in use, including what arrived embedded in the LMS, SIS, CRM, and advising software nobody separately approved? 3. Authority. Has the Human Authority Line™ been drawn and documented for each high-risk system: the point where AI's recommendation ends and a person's judgment must remain? Article 14 of the EU AI Act sets the same bar for high-risk systems, that a person must be able to effectively oversee and override the output, and it is a useful reference point even outside EU jurisdiction because it gives oversight a specific, demonstrable meaning. 4. Reputation. If an AI system failed publicly, is there a named person already authorized to act immediately, with the response documented in advance? 5. Design. Once built, does your own team fully own and operate this architecture?
Falkovia's point of view
Presidents already recognize AI as the defining force facing their institutions. The gap sits in the human architecture built underneath it, and that gap is where governance is won or lost. I have said this before and the 2026 data keeps proving it true: AI adoption is ten percent technology and ninety percent human architecture. In practice, that means naming who is accountable for each AI system, telling your team plainly where their judgment still overrides the machine, and rehearsing your response to a public AI failure before one happens.
A named owner is only the starting point for that accountability. The harder question is whether that person is actually accountable for what the AI decides, holds real authority to override it, and answers for the outcome when it is wrong. Leading differently right now means auditing AI with the same regularity boards already apply to finance and risk: where AI lives across the institution, who is accountable for what it produces, and whether data on overrides, how often they happen and whether people feel safe making them, actually reaches the board table.
Ahead of moderating this exact question at the IHE Summit this October, my answer is this: being AI-informed is the easy part. Proving your institution has built the human architecture underneath its AI is the harder and more consequential test of presidential leadership right now. The timing is exact: Deloitte's own data shows agentic AI adoption is projected to more than triple within two years, from 23 percent to 74 percent of organizations. That is three times as many systems, moving faster and with more agents involved, and the architecture to govern them is far easier to build now than to retrofit once the speed and the number of agents have both multiplied. See the full accreditor framework