Universities invest in AI, but staff and students bypass institutional tools
Universities are spending heavily on artificial intelligence systems, yet educators and students routinely ignore them in favor of tools outside institutional control. The problem isn't technology-it's trust.
Researchers at the University of Warwick studied how staff and students actually use AI and found a striking pattern: approved institutional tools sit unused while "shadow AI"-unapproved software-proliferates. This disconnect reveals a governance failure, not a technical one.
The issue centers on a four-party relationship: students, educators, AI systems, and institutional governance. When any party lacks trust in the others, adoption stalls.
Why the trust breaks down
Universities typically prioritize integrity when adopting AI-focusing on risk control, data security, and rule enforcement. This often means restricting tool access or mandating systems with limited capabilities.
An educator who finds a tool that significantly improves teaching but faces institutional restrictions faces a choice: follow the rules or use shadow AI. When the approved tool underperforms compared to alternatives, educators question whether the institution prioritizes their success or merely compliance.
This creates what researchers call a "benevolence gap." The institution's safety measures feel like barriers rather than support, so educators place more trust in shadow tools than in institutional gatekeepers.
Building institutional trust requires three shifts
Be transparent about AI use. Educators must discuss AI openly with students, even when using non-approved tools. Design assignments where AI's role is clear and focus assessment on judgment and critique rather than first-draft production alone.
Prioritize capability over compliance. Choose tools that genuinely support teaching and learning. Institutions should ask educators for examples of real workflows-drafting, coding, feedback generation, literature research-and select tools that handle these tasks well.
Reduce friction in approval processes. Limited access, complex sign-offs, inconsistent guidance, and platforms that lag behind public alternatives all signal that using AI requires workarounds. This normalizes shadow practices. Instead, make approved tools easily accessible, functionally competitive, and supported with clear guidance.
Allow adoption at different speeds
Institutions should offer adoption as a choice, not a mandate. Some educators experiment early; others prefer cautious observation. Support both approaches through training, shared examples of practice, and sandbox environments where staff can experiment safely.
These testing environments are standard in industry and regulatory settings but remain uncommon in higher education despite their proven value.
The shift ahead
Universities must move beyond regulation to building trust conditions. The goal isn't to control AI use but to enable educators and students to decide how, when, and whether to use it.
For education professionals, this means your institution's AI policy will only succeed if it treats you as competent, supports your work, and follows through on its principles consistently. If it doesn't, you'll keep using shadow tools-and the institution will keep wondering why adoption stalls.
Learn more about AI for Education and explore practical approaches to implementing AI in your teaching through our AI Learning Path for Teachers.
Your membership also unlocks:
