Artificial intelligence tools are becoming part of everyday administrative and operational work across higher education. When used thoughtfully, AI can help Northern State University employees organize information, draft communications, summarize materials, analyze patterns, and improve workflow efficiency. This guide is intended to help administrative and operational users understand appropriate, responsible, and policy-aligned ways to use AI in their work. It emphasizes practical use, data protection, human review, and professional accountability so that AI supports, not replaces, the judgment, expertise, and responsibility of NSU employees.

This page offers practical guidance on when AI is an appropriate assistant for office and administrative tasks, and when it is not.

 

Frequently Asked Questions

Which AI tool can I use with this data?

Using Microsoft 365 Copilot (signed as with NSU credentials) or BOR-approved AWS Bedrock?

✅ Public Data
✅ Internal Data
✅ Institutional Data
✅ Regulated Data
✅ Restricted Data
 

Using ChatGPT, Claude, Gemini, Perplexity, NotebookLM, Canva, or other public AI tools?
 

✅ Public Data
✅ De-identified Data
✅ Aggregated Data
❌ Regulated Data
❌ Restricted Data
❌ FERPA Records
❌ HR Records
❌ Protected Financial Data
❌ Health Information
❌ Internal Institutional Data (i.e. meeting notes, internal documents, etc.)
 

When in doubt, do not upload the information and consult Technology Services or your supervisor before proceeding.
 

BOR and NSU Approved Guarded AI Tools
 

These tools operate within approved institutional environments and may be used with university data, including Internal, Regulated, and Restricted information when used in accordance with Board and university policies
 

What data can I use in an approved guarded AI environment?

Approved guarded AI environments may be used with:

  • Public data
  • Internal data
  • Institutional data
  • Regulated data
  • Restricted data

Examples include:

  • FERPA-protected student information
  • Student IDs and academic records
  • Personnel and HR records
  • Protected financial information
  • Health-related information
  • Other protected institutional records

Employees should still follow all applicable university policies, data classification requirements, and access controls when using these tools. Protected data should be anonymized whenever practical and only shared on a need-to-know basis
 

What can I do with approved guarded AI tools?

Approved tools may be used to:

  • Draft emails, reports, and communications
  • Summarize meeting notes and documents
  • Analyze institutional data
  • Organize and categorize information
  • Create presentations
  • Identify trends and patterns
  • Assist with administrative and operational work

Employees remain responsible for reviewing all output, verifying facts, and exercising professional judgment before using AI-generated content.
 

What are public AI tools?

Public AI tools include services such as:

  • ChatGPT
  • Claude
  • Google Gemini
  • Perplexity
  • NotebookLM
  • Canva Magic Studio
  • Personal AI subscriptions and accounts

These tools may be useful for brainstorming, drafting, research, summarization, and general productivity work, but they are not approved for protected university data.
 

What information can I put into a public AI tool?

Public AI tools may be used with:

  • Public information
  • Information already available on public websites
  • De-identified information
  • Aggregated information
  • General examples
  • Anonymous content

Examples:

✅ Asking AI to improve wording on a public webpage.
✅ Asking AI to draft an event announcement using publicly available information.
✅ Uploading aggregated statistics with no identifying information.
 

What information should never be entered into a public AI tool?

Never enter Regulated, Restricted, or internal university information into a public AI platform. 
Examples include:

  • FERPA-protected student records
  • Student names and ID numbers
  • Grades and advising records
  • Employee personnel files
  • Confidential HR information
  • Protected financial information
  • Health information
  • Social Security numbers
  • Unpublished sensitive research
  • Other protected institutional information
     

Does a paid subscription make a public AI tool safe for university data?

No.

A personal or paid subscription may provide additional privacy controls, and some vendors may allow training to be disabled. However, university data classification requirements still apply. A paid account does not automatically make a public AI service an approved environment for regulated or restricted university data.
 

Can AI make decisions for me?

No. Although AI tools may be used in a variety of ways, professional judgment always remains the responsibility of the employee using the AI tool. 

What AI Can Do in an Office

  • Summarize notes: Turn long meeting recordings into short action lists.
  • Draft documents: Write first drafts of reports, emails, or job postings.
  • Search data: Find internal files, policies, or past project notes instantly.
  • Sort information: Categorize customer support tickets or parse messy spreadsheets.
  • Translate text: Convert emails and messages into multiple languages quickly
     

What AI Can't Do in an Office?

  • Feel empathy: Handle sensitive workplace conflicts, grief, or emotional support.
  • Take blame: Accept legal, ethical, or personal responsibility for a failed project or decision.
  • Read nuance: Understand unspoken office politics, hidden team dynamics, or unwritten rules.
  • Initiate goals: Know what problem to solve next without a human setting the direction.
  • Build deep trust: Form authentic mentor-mentee relationships or inspire team culture.
     

Can I use AI to draft emails, reports, presentations, or meeting notes?

Yes, but think of AI-generated drafts as starting points rather than finished products.
These are excellent examples of appropriate AI use when the information being shared is appropriate for the AI platform being used.

  • Emails
    • Write a short email for a new employee
    • Turn a brief note into a polite reminder
    • Answer common support questions fast
    • Change a harsh draft into a polite one
  • Reports
    • Turn raw numbers into a clear paragraph
    • Condense a long document into key points
    • List of your office’s top progress items
    • Pull and sort main facts from raw notes
  • Presentations
    • Make a side-by-side plan for a talk
    • Cut long paragraphs into short slide text
    • Write what to say for each slide
    • Suggest charts or images for data
  • Meeting Notes
    • Remove filler words from a live recording
    • Pull out a clean list of who must do what
    • List choices made during the call
    • Write a summary to send to absent co-workers
       

Always:

  • review the draft carefully
  • personalize it as needed
  • verify all facts
  • ensure the final document reflects your own professional judgment
     

Can AI help me analyze data?

Yes, but remember that protected data should only be used within approved Board or university AI environments and anonymized whenever possible.

AI can be extremely helpful for:

  • identifying trends
  • comparing information across multiple datasets
  • organizing findings
  • summarizing results
  • highlighting possible patterns
  • generating ideas for further investigation

However, AI should not be considered the final authority on data interpretation.
Human expertise remains essential to:

  • verify findings
  • interpret results
  • determine significance
  • identify errors or incorrect assumptions
  • communicate appropriate conclusions
  • What information should never be entered into a public AI tool?

Never enter Regulated or Restricted institutional data into public AI systems.
Examples of regulated or restricted data can be found in BOR policy 7.9 or the data governance and security page for more information on what data can be used with certain AI tools.

Whenever possible, use:

  • anonymized information
  • aggregated summaries
  • generalized examples
  • de-identified data
     

How accurate is AI?

AI can be extremely helpful, but it has limitations and is not always correct. Every AI-generated output should be reviewed, edited, fact-checked, verified before being used for official university purposes. 

AI may misunderstand context, oversimplify complex situations, make incorrect assumptions, omit important information, and/or generate inaccurate or fabricated content.
 

Can AI create official university documents for me?

AI may assist in creating first drafts, but it should not create final documents without human review. AI supports documentation; people remain accountable for it.
Think of AI as a capable first-draft assistant, not the author of official university documents. Every official document must be reviewed, validated, and approved by the responsible employee before it is shared or published.

The employee is responsible for:

  • accuracy
  • completeness
  • context
  • tone
  • compliance
  • final approval

Meeting Minutes - AI summarizes notes from a Council meeting.

  • Can you post the AI summary as the official minutes?
    • Answer: No.
  • You should:
    • Verify motions, votes, and decisions. 
    • Add missing context or clarifications. 
    • Ensure the summary accurately reflects the meeting. 
    • Approve the minutes before distribution. 
  • Key Point: AI may organize notes, but the official record requires human verification. 
     

Can AI generate evidence, reports, or outcomes that did not actually occur?

AI should help communicate actual work, not manufacture it. 

AI should never be used to:

  • fabricate evidence
  • invent data
  • create fictional outcomes
  • misrepresent university activities
  • misrepresent authorship
     

Do I need to disclose that I used AI?

Yes, when appropriate. Transparency builds trust and supports professional integrity.
Employees should follow university and Board guidance regarding disclosure of AI-assisted operational work. Regardless of disclosure requirements, employees remain responsible for the final work product.

Examples:

Internal Email - You use AI to help draft an email reminding students about registration deadlines. You review the message, verify all dates, personalize it, and send it.

  • Question: Should you disclose that AI helped draft the email?
  • Answer: Generally, no if AI was used as a writing assistant, and you reviewed and approved the final message.
  • Key Point: Employees remain responsible for the content, even when AI assists with drafting.

Administrative Report - You use AI to help summarize themes from student satisfaction survey comments before presenting the report to university leadership.

  • Question: Should you disclose that AI assisted with organizing or summarizing the comments?
  • Answer: Yes. Transparency helps readers understand how the report was prepared and supports confidence in the process.
  • Key Point: AI can assist with analysis, but the employee validates the findings and is responsible for the conclusions.
     

How should I incorporate AI into my normal workflow?

AI should make people more effective; not replace the expertise they bring to their work. 

A recommended workflow is:

  • Define the task or problem.
  • Determine whether the information is appropriate for the AI platform.
  • Use AI to organize, draft, summarize, or analyze.
  • Review the output critically.
  • Correct errors and provide additional context.
  • Apply your professional judgment.
  • Approve the final product before sharing or using it.
     

 

Resources

Work Smarter, Not Harder: 10 Benefits of AI in Your Workplace
https://ai.utah.edu/innovate.php
https://www.cardinaleducation.com/how-ai-is-changing-the-college-admissions-process/ 
Northwestern University Information Technology