The University AI Didn''t Replace: Rethinking Universities in the AI Era

Synthesis: Rather than replacing universities, generative AI redefines their essential functions — this paper proposes a four-level framework of institutional AI adoption and argues that the central challenge is moving from isolated, individual-driven experimentation to strategic integration, supported by workload and recognition systems.

Core Argument

The paper positions AI not as an existential threat to universities but as a catalyst for reimagining what universities do. Drawing on a case study of AI-enabled curriculum initiatives across several units at one institution, the authors contend that most universities remain in the early stages of adoption, where AI innovation occurs informally and without institutional recognition. The key institutional challenge is therefore not technological provision but alignment: redesigning learning around AI-supported reasoning and aligning policies, workload models, and recognition systems to support educational transformation. This connects to broader discussions about the frameworks needed for higher education, and complements the finding that most staff already use AI for work.

The Four Levels of AI Adoption

The paper's central framework describes four stages of institutional response, ranging from defensive restriction to full integration, noting that many institutions presently operate around Level 0 or Level 1:

1. Level 0 — Defensive Containment: AI is treated primarily as a threat to academic integrity. Responses focus on restricting use (discouraged or banned in assessment, heavy reliance on invigilated exams) with little or no curriculum redesign — often creating a mismatch between policy and the reality of widespread student AI use.
2. Level 1 — Informal or Peripheral Adoption: Experimentation happens largely at the level of individual educators rather than through institutional strategy, with limited or inconsistent guidance for students, no formal recognition in workload or promotion, and uneven adoption across units. Innovation thus depends on motivated individuals and can burden those leading change.
3. Level 2 — Strategic Integration: Universities begin embedding AI into teaching policy and curriculum design, with clear institutional principles, curriculum redesign for AI-rich environments, professional AI development for educators, and recognition of innovation in workload or funding. Teaching increasingly focuses on reasoning, interpretation, and responsible use.
4. Level 3 — AI-Embedded or Transformational Universities: AI becomes part of core educational infrastructure and the learning ecosystem — integrated into the LMS, assessment, and learning support; curricula centered on AI-supported reasoning rather than content delivery; assessments that evaluate judgment and justification in AI-rich environments; and students working with AI as a cognitive partner in authentic Problem Solving.

Key Strategic Steps

To move beyond Level 1, the paper recommends concrete institutional actions: recognize AI-driven curriculum redesign in workload models so innovation is not dependent on unrecognized effort; embed AI-enabled teaching innovation in promotion and teaching-award criteria; establish clear institutional principles for AI use in learning and assessment; redesign assessment toward reasoning and justification in AI-rich environments rather than merely tolerating AI; and create institutional pilots or funded initiatives that scale successful innovations beyond individual courses.

Connections to Knowledge Base

• Shares the institutional perspective with on human/institutional capacity bottlenecks
• Contrasts with the Guidelines for Designing AI Technologies to Support Adult Learning focus on learner-facing technology guidelines
• Extends Teacher AI Competency from individual educators to institutional competency — institutional readiness is a distinct level above individual educator skill
• The workload-and-recognition emphasis speaks to Educational Development centers navigating GenAI adoption, and the assessment implications echo Assessment Validity concerns about what assessment means when AI can produce university-level work
• Relevant to AI Regulation in Education discussions about higher-education policy and A principled way to think about AI in education: guidance for educators and policy makers based on goals, models frameworks

Open Questions

• How do different national contexts (US, EU, Global South) shape university AI responses?
• What is the timeline for meaningful institutional transformation vs. superficial adoption?
• How does institutional rethinking interact with A principled way to think about AI in education: guidance for educators and policy makers based on goals, models frameworks?
• Which structural incentives most effectively convert informal educator experimentation into scaled, recognized innovation?

What this means for practice

• Administrators. Locate your institution on the four-level framework before funding more tools: many institutions currently operate around Level 0 or Level 1, where experimentation depends on motivated individual educators and is unrecognized in workload or promotion structures.
• Administrators. Recognize AI-driven curriculum redesign in workload models and embed AI-enabled teaching innovation in promotion and teaching award criteria, so that scaling does not depend on unrecognized effort by individual educators.
• Instructors. Redesign assessment toward reasoning and justification rather than trying to contain AI: the case units use live pitches, reflective portfolios, and viva voce to make students justify, explain, and adapt their thinking in real time.
• Faculty developers. Build shared sensemaking about change rather than assuming that new tools or individual experimentation will suffice; the paper cites Kezar (2018) on change efforts that falter under implicit and overly simplistic theories of change.
• Administrators. Create institutional pilots or funded initiatives that scale successful teaching innovations beyond individual courses, since isolated innovations otherwise remain unrecognized and unable to spread.

Limitations

• The framework is illustrated through a small, purposively chosen set of cases — AI-enabled curriculum initiatives in several units at the authors' own institution, plus published accounts of a Macquarie University unit and a Nanjing Normal University unit delivered to over 250 undergraduate students — rather than a systematic or representative sample of institutions.
• The authors state that it is not completely clear whether the flagship "AI-native" examples (LSI and SUTD) are genuinely at the strategic adoption stage, so level assignments rest on public positioning rather than verified practice.
• No outcome data on student learning, or on the effects of the recommended workload and promotion changes, are reported; the strategic steps are recommendations to be tested rather than evaluated interventions.
• Case evidence spans Australia, the UK/Singapore, and China, and the paper notes that national policy levers such as TEQSA guidance may push adoption toward Level 2 or toward fragmented, ad hoc responses — a contingency the framework does not model.

Citation

Binkowski, K. P., & Hopkins, A. (2026). The University AI Didn't Replace: Rethinking Universities in the AI Era.