Guidance
Plain and practical, for staff.
Where AI fits in staff work
A lot of the work you do every day is exactly what these tools are good at: drafting, summarizing, organizing, and turning rough material into something finished. The point of this page is for you to see some guidance for the type of work you do. The same data rules from the For everyone page apply to everything below: sign in with your NetID, and match the tool to the sensitivity of the information.
These aren't everything possible, just starting points. Select any one to see examples.
Writing and communications Draft it faster, then make it yours.
First drafts of emails, memos, announcements, and talking points. Tightening something you've already written, adjusting tone for a different audience, or turning bullet points into prose. Translating a dense policy into plain language for the people who have to follow it.
Documents and reports Structure, summarize, and format.
Outlining a report before you write it, summarizing a long document or a stack of meeting notes, building a first-draft template, or cleaning up formatting and consistency across a file. Drafting standard operating procedures and FAQs from how you actually do the work. Outlining and building a PowerPoint deck from material you upload or research it pulls together for you.
Spreadsheets and data Formulas, lists, and budgets; check the math.
Writing or explaining a formula, reorganizing or de-duplicating a list, drafting a chart, or describing what a dataset shows. Building a new grant budget from a template or a prior year's, working from the proposal or narrative. Keep university data in MSU tools, and remember that AI gets numbers wrong with confidence: check the math.
Scheduling and workflow Take the busywork off your plate.
Drafting agendas, organizing project tasks, summarizing where a project stands, or proposing a meeting structure. Turning a long thread into a clear set of next steps. Finding open times across several colleagues' calendars and setting up the meeting.
Service and support Answer common questions consistently.
Drafting answers to routine questions, building knowledge-base content, or preparing onboarding material. Public-facing AI assistants are a different matter and need approval before they're built and before they go live: start that with the Chief Information Security Officer at security@msstate.edu.
Meetings Capture it, but use the right tool.
Use Microsoft Teams or Webex for recording and transcription. Do not use Read.AI to record or transcribe meetings; if it's on your computer and you can't remove it, submit a ticket through the ITS Service Desk.
How much should AI decide?
Not every task needs the same amount of you. Some can run almost on their own; others should never leave your hands until you've made the call. A simple way to think about it: the more a task affects a person, a record, or the public, the more human judgment it needs before anything is final.
This is the staff version of a principle that runs through all of MSU's guidance: a person is accountable for the work, whether or not AI helped.
| Tier | What it means | Where it fits |
|---|---|---|
| 1 · Run it | AI does the task start to finish; you check the result. Low-stakes, routine, easy to undo. | Formatting a document, drafting an agenda, cleaning a list, summarizing your own notes, turning bullets into prose. |
| 2 · Draft it, you approve | AI produces; nothing is sent, posted, or filed until you review and approve it. | First-draft emails and announcements, FAQ answers, newsletter copy, meeting summaries, routine reports, a first-draft budget based on last year's. |
| 3 · Inform, don't decide | For consequential matters, AI can gather, summarize, or lay out options, but a person makes the call and owns it. | Summarizing vendor quotes for a recommendation, organizing materials for a committee, drafting analysis behind a budget decision. |
| 4 · People decide about people | The decision stays with a person. AI can help with the work around it (comparing candidates against a set rubric, or drafting an evaluation summary you then own) but it does not rank people or make the call. | Hiring, performance evaluation, discipline, promotion, accommodations, and financial aid determinations. |
The hard line. In Tier 4, the decision stays with a person, full stop. AI can help with the work around it: comparing candidates against a set rubric, organizing materials, or drafting an evaluation summary you then own and verify. What it must not do is rank people or make the call. Employment decisions carry real exposure and real bias risk, and the accountability for them belongs to a person. When you're unsure which tier a task lands in, treat it as the higher one and ask.
Disclose your use
Disclosure is a best practice. When you share work that AI meaningfully helped with, say so.
Transparency builds trust. People know what they're looking at and who stands behind it. Just as useful, it shows your colleagues that these tools are part of normal, capable work, and it gives them a concrete picture of how AI might help with their own tasks. The goal is to make disclosure ordinary and unremarkable. It's not a confession, or a sign you don't know how to do something. You always own the output.
A few practical notes:
- Scale it to the work. A quick grammar pass or a formatting cleanup doesn't need a label. Work where AI shaped the substance (a drafted report, an analysis, a slide deck) should say so. In research, follow the norms of your discipline and your publisher.
- The judgment is still yours. Disclosing AI use doesn't move responsibility onto the tool. You verify the facts, the names, the numbers, and the tone, and you own the result.
Attribution that fits the work
A disclosure can be one line in a footer, an email sign-off, or a note on a slide. The right amount of detail depends on how much AI did and how much you did. Keep it honest and keep it simple.
Light touch, for routine assistance
- Assisted by Copilot
More specific, when you want to show the division of labor
- Assisted by Claude, reviewed and approved by [your initials]
- Outlined by ChatGPT, written by me
- Research and drafted by Copilot, reviewed and approved by [your initials]
Explain how the tool helped, and make clear that you reviewed and stand behind the output. A statement like this works on a document footer, the first slide of a deck, the bottom of a web page, or the close of an email.
Working with people's data
Staff work touches some of the most sensitive information at the university: personnel records, applicant materials, student records, financial and benefits data. That doesn't put AI off-limits; it tells you which tool to use and when to ask first.
The full rules, with examples, are on the For everyone page. The short version for staff:
- Personnel and non-public student records are MSU Moderate Risk. They go only into MSU tools you're signed in to with your NetID. A non-MSU tool needs approval first.
- SSNs, financial accounts, health information, and similar are MSU High Risk. Identifiable High Risk data goes into no AI tool at all. De-identification is possible, but discuss it with the CISO first at security@msstate.edu.
- Applicant and reference material deserves extra care. It's tied to a decision about a person (Tier 4 above) and often falls under Moderate or High Risk. Keep it in MSU tools, and keep the ranking and the decision with a person.
If you wouldn't email it outside MSU, don't paste it into a tool outside MSU.
Data-classification questions go to the data steward for the system that owns the information (HR for personnel, the Registrar for student records), and anything else you're unsure about goes to the Chief Information Security Officer at security@msstate.edu.
The bottom line
AI is part of the daily work of staff across Mississippi State: in communications, documents, scheduling, service, and reporting. Used well, it saves time and gives you space for higher-value tasks.
This is advisory guidance, not formal university policy. It does not require anyone to use AI; choosing not to is a completely legitimate choice, and one we respect. It is written for all faculty, staff, and students, in every context where AI is used with university information, whether the tool is free, paid personally, or paid by the institution.
It is consistent with MSU's existing policies, including OP 01.10 (Information Security Policy) and OP 01.12 (Use of Information Technology Resources). Where this guidance and existing policy address the same matter, the existing policy controls.
Looking for something else? Faculty have their own version on the Faculty page, students on the Students page, and general guidance for everyone at MSU on the For everyone page. For MSU's full Information Security Program, see infosecurity.msstate.edu.
Last updated June 19, 2026 · Mississippi State University AI Hub