The CIO’s starter guide to leading AI adoption on campus

Starter guide checklist:

• Define AI principles (e.g., embracing AI as a tool for learning, committing to digital inclusion) to guide institution-wide adoption
• Form governance structures with leadership oversight and implementation teams that designate accountability for mitigating AI risks and exploring opportunities
• Position faculty in general governance bodies and/or convene specific subcommittees to help guide ethical use of AI (e.g., reviewing AI-enabled solutions)
• Formalize protocols to fast-track review of low-risk AI projects while establishing escalation paths for higher-risk projects
• Centrally track AI projects and applications to curtail shadow AI projects and ensure all efforts abide by institutional risk and vendor criteria

Starter guide checklist:

• Prioritize IT AI staff roles based on institutional AI posture (e.g., data science skill sets when building solutions vs. staff with knowledge of institutional systems when buying)
• Educate senior leaders about AI cost realities and the need to diversify funding sources for sustainability (e.g., budget reallocation, student fees)
• Move beyond general awareness and instead build tool- and role-specific training to help users integrate AI tools into their workflows
• Advertise IT’s role as a consultative partner for campus AI adoption and experimentation (e.g., hosting symposiums, office hours)

Starter guide checklist:

• Design right-fit pilots for campus to accurately evaluate applications and maximize ROI; recommended practices include:
  • Showcase solutions to senior leaders first
  • Select units/roles with most apparent business need (and empower managers to decide)
  • Train users early and continuously
  • Survey use to gauge efficiency gains
  • Build a Community of Practice to collaborate on use cases
• Assess application risk (i.e., input and output data, complexity of the role AI fulfills) first to inform build versus buy decisions
• Match data preparation to application risk instead of trying to fix all institutional data at once (e.g., a staff FAQ chatbot requires less prep than a one-stop student assistant)
• Partner with functional experts to test applications, ensuring they address relevant problems accurately