Iowa funds supported AI experimentation, including agentic tools

Discuss the token-program model with campus IT, administrative service owners and finance. Ask which workflows need experimentation, how access requests will be prioritized, and what funding supports successful tools after exploration. A bounded engagement could design an intake and evaluation process for one department. The value hypothesis is controlled learning about useful automation, with a clear stop decision for weak candidates. This announcement supports discovery about operating readiness, not an inference that Iowa seeks a supplier or that allocated funding is available to a particular engagement. Do not promise production readiness or savings from access alone.

For agentic experimentation, provide isolated repositories, synthetic administrative records, restricted outbound connections and explicit execution approval. Prerequisites include documented owners, a test harness and a supported identity path. Integrate usage accounting without copying sensitive prompts into broadly accessible logs. Validate a representative workflow against the existing process, testing misleading retrieved instructions, budget exhaustion and attempted unauthorized writes. Proposed acceptance should include blocked out-of-scope actions and reproducible outputs that pass domain review. Keep production promotion as a separate decision because the announcement provides no tested architecture or evidence that autonomous actions are safe.

Build intake, budget allocation, onboarding, incident handling and retirement procedures under an ITS service owner. Department leads should supply use cases, reviewers and a current-process baseline; security should approve data and permissions. Train participants to recognize fabricated output and unsafe execution requests. Proposed acceptance includes a named owner and data classification for every trial, functioning spend limits, demonstrated rollback for permitted actions and a documented continue-or-stop review. Risks include experiments becoming unsupported production services, uneven staff access and recurring costs exceeding the exploration budget. Measure service quality and total effort before any expansion.