Higher education IT and research leaders are being reminded that AI projects fail not because of models but because underlying institutional data is fragmented, inconsistent and poorly governed. The piece argues universities must invest in data stewardship and governance before deploying generative or predictive systems. The reporting lays out a practical roadmap: catalog data sources, set clear ownership and lineage, and implement operational controls so AI outputs are auditable and trusted by faculty and administrators. For readers: "data governance" means policies and technical controls that ensure data is accurate, traceable and used ethically across systems. Chief information officers and deans should see this as a call to reprioritize data architecture spending, align research‑data management with institutional risk frameworks, and integrate compliance work (FERPA, export controls) into AI rollouts to avoid reputational and legal exposure.