University Supervisors' Meeting: Sept 17 Resources

Return to Supervisor Network

Meeting Topic

AI Tools for Supervisors

Meeting Resources:

Meeting Summary:

September's session centered on responsible AI use, with Dr. Carlos Lizárraga, a data scientist manager with the Arizona Institute for AI and Society and the Office of Responsible AI, walking supervisors through UA's approved AI tools and how to use them safely. He framed the session around three data classification levels — public, protected, and restricted — and matched each to specific tools: Microsoft Copilot and Google Gemini for public and protected work, and the UA GenAI Platform as the only option for restricted data covered by FERPA, HIPAA, or similar protections. He also introduced two prompt-writing frameworks, RACE (Role, Action, Context, Expectation) and RISEN (Role, Instructions, Steps, End goal, Narrowing), and walked through real examples supervisors can adapt for tasks like performance narratives, job postings, and meeting agendas.

A live poll showed supervisors split fairly evenly on their AI comfort level: 33% said they're ready to use it, 34% called themselves cautiously curious, 21% are still figuring it out, and 11% asked to be checked back with in six months. Q&A discussion focused on practical edge cases — whether UA Global Campus staff have the same Copilot access despite logging in differently, and how to carry prior work forward for supervisors who'd already been using personal AI accounts before the GenAI Platform existed (Dr. Lizárraga noted the platform currently starts every chat blank, with no memory of past sessions). Several supervisors also asked for clarity on where protected versus restricted data lines fall, prompting a reminder that Gemini and Copilot are fine for protected information but off-limits for anything FERPA-, HIPAA-, or PII-governed. Dr. Lizárraga confirmed a HIPAA-compliant version of the GenAI Platform is in development and expected within the next couple of months, closing a current gap where health-related data has nowhere fully compliant to go.

The meeting closed with a segment from the CaT Spokes community — Michael Dorland, Kat Salthouse, Karren Fultz, and Nikki Pikaard — introducing their cross-campus communities of practice, including AI, Microsoft, and process automation office hours, the Leadership Learning Network, and the Project Management Community Academy, all open to staff, faculty, and students looking for ongoing peer support beyond this monthly series. 

Key Takeaways:

  • U of A's three data tiers (public, protected, restricted) determine which AI tool is allowed; when in doubt, restricted-level tools (UA GenAI Platform) are always safe to default to.
  • The UA GenAI Platform is the only approved tool for FERPA, HIPAA, or PII-governed data; Copilot and Gemini are approved for protected but not restricted information.
  • RACE and RISEN give supervisors a repeatable structure for prompts, specifying role, action, context, and expected output consistently improves results.
  • Every AI output needs human review before use; the tools can produce confident-sounding but inaccurate information, especially with names, dates, and policy references.
  • A HIPAA-compliant GenAI Platform option is coming within the next couple of months, closing the current gap for health-related employee data.

Action Steps:

  • Classify your data (public, protected, or restricted) before choosing a tool for any AI-assisted task.
  • Strip names, IDs, and other identifying details from prompts, replacing them with role placeholders like "the employee" or "the applicant."
  • Try RACE or RISEN on your next drafting task (an email, agenda, or job posting) to see the difference structured prompting makes.
  • Start a shared prompt library for your team — even three saved prompts is a useful starting point.
  • Explore Cat Spokes office hours (AI, Microsoft, or process automation) if you want hands-on help applying any of this.

Next Meeting: Oct. 15, 2026: Talent Acquisition Updates