Employees performing HR tasks throughout the University of Minnesota can use generative AI to improve service for fellow employees, reduce risk, and learn together.
This guidance was created to help employees who perform HR tasks to use AI tools safely and effectively, consistent with University policy, data classifications, and our values of equity and transparency. With the fast-paced changes in technology and the legal landscape, please check back often for updated guidance, resources and support on the responsible use of AI in the human resources tasks and functions.
Getting Started
Use UMN‑licensed tools
- Use only the Approved Tools when at all possible. Do not use unapproved tools without review by OHR or OIT. Learn more about UMN OIT’s review process.
- Coordinate with your OIT Lead to work with UMN OIT to pilot tools to prevent shadow systems and protect data and other systems.
Think data first
- Consider what data the tool actually needs and provide the minimum into approved tools.
- Put only Public data into AI. Minimize or de-identify data if necessary.
- Never paste Private‑Restricted or Private‑Highly Restricted data into unapproved tools. Visit the Know Your Data and How to Protect University Data webpage to check data privacy levels.
Check bias
- Review prompts and outputs for stereotypes, exclusionary wording, or inconsistent criteria; use inclusive language guides; escalate concerns to OHR or the Office of Equity and Diversity. Learn more about unconscious bias and reach out to our EDI team for guidance.
Own the process and the outcomes
- AI should never make an HR decision for any process. AI can save time and provide many different kinds of support, but a human must always review what AI creates or outputs and modify it before sharing.
HR use cases and risk factors by function
While this does not encompass all possible use cases, the examples below illustrate how AI might be integrated into a team's workflow, and how varying levels of risk may influence the need for caution and approval. Consult your HR or IT leads for any further guidance and clarification.
Human review is always required. For Medium and High-risk activities, document how you reviewed and approved the final product.
Talent Acquisition (TA)
Low Risk
- Interview Questions: Use AI to create competency-based questions for faculty, staff, and student roles. HR reviews all questions. No candidate personally identifiable information (PII) should be included.
- AI Disclosure: Document and disclose any meaningful AI assistance used in hiring or decision-making files.
Medium Risk
- Job Description Drafting: Auto-generate job description language aligned with classification standards. HR must validate content before posting. Minimum qualifications and salary details must remain unchanged. No applicant data should be used in prompts.
- Job Announcements: Generate job announcements based on validated job descriptions.
- Job Competencies: AI may suggest competencies for roles. HR must review for accuracy.
High Risk
- Activities related to applicants of any kind. All applicant data is classified as private-restricted, and should be treated accordingly when using AI tools.
- Candidate Screening: Do not use AI to screen or rank applicants or resumes.
- Interview Assessment: If using asynchronous interview tools, ensure that responses are reviewed by humans to mitigate bias.
- It is not recommended to use Zoom AI Companion during candidate interviews.
- If it is deemed necessary to use the AI Companion features, the host must verbally notify the candidate at the start of the interview and explain how the transcript will be used.
- Please make sure to schedule separate debrief meetings with the hiring committee; do not debrief in the same session as the interview..
- Exercise caution to make sure meeting transcripts are only sent to the owner of the meeting, not the entire hiring committee.
Total Rewards
Low Risk
- Communication Templates: Draft FAQs for retirement, tuition, and wellness programs. HR validates accuracy and links before publishing.
- Comparison Summaries: Create plain-language summaries of benefit plan options based on official documents. HR validates before publishing.
Medium Risk
- Job Description Drafting: Draft job description text and summarize role responsibilities. Managers and employees are accountable for accuracy. Compensation analyst validates.
Employee & Labor Relations (ELR)
Low Risk
- Policy Interpretation: Draft plain-language summaries of public benefits information. Link to official sources. HR validates before publishing.
Medium Risk
- Scenario-Based Guidance: Generate sample scripts for handling difficult conversations. HR/ELR/manager must review. Prompts must not include real case details.
High Risk
- Eligibility Determination: Do not use AI to determine employee eligibility.
Strategy & Workforce Planning
Low Risk
- Environmental Scans: Summarize public HR trends relevant to higher education. Cite and verify sources.
- Strategic Briefs: Link national trends (e.g., AI, equity, flexible work) to the UMN context. HR leadership must review.
- Benchmarking: Compare public peer reports. Document sources and assumptions. Ensure accuracy.
Medium Risk
- Regulation Summaries: Summarize new laws and their implications for higher-ed HR. Cite and validate sources. HR and OGC must review before publishing.
Performance & Development
Low Risk
- Development Plan Content. Ideate on actionable strategies to develop specific skills required for their role. Manager, SMEs or HR review and cite sources.
Medium Risk
- Performance Documentation Content. Collate or summarize any performance documentation, feedback and/or recognition to include within the performance review or performance improvement document. Pro tip: remove references to specific employee identifiable information (ie, name) whenever possible. The manager is accountable for all performance documentation and ensuring it is 100% accurate and reflective of the employee's performance.
Learning & Development
Low Risk
- Training Content: Create outlines, job aids, knowledge articles, and microlearning materials. HR and SMEs review and cite sources.
- Learning Summaries: Condense emails and training materials into key takeaways. Include citations and links. HR and SMEs validate before publishing.
- Assessment Tools: Draft surveys, learning interactions, reflection prompts, and evaluation tools. Summarize results without sharing raw personally identifiable information (PII). HR reviews before publishing or acting on results.
Medium Risk
- Personalized Learning Paths: Suggest learning paths based on roles and goals using approved tools and high-quality sources. No performance data or personally identifiable information (PII) should be uploaded. HR and managers validate before sharing.
HR Analytics & Workforce Data
Medium Risk
- Narrative Summaries: Summarize public dashboards or de-identified data samples. HR validates before sharing.
- Audit Preparation: Summarize gaps and risk exposure using approved data sources and architecture. Do not use restricted data. HR validates before and after using AI tools.
High Risk
- Predictive Analytics: Do not use AI to model or make predictions based on HR data. Instead, you can use AI as a consultative tool to tell you which metrics might be important for things you want to predict (e.g. turnover intention), and use approved reporting and analytics to see those metrics.
Approved tools (UMN-licensed)
Use these tools for University work as they provide enterprise‑grade protections. Always follow the tool‑specific guidance and review UMN OIT's guidance before proceeding
- Gemini (Enterprise‑Grade Data Protection)
- Copilot (Enterprise Data Protection)
- NotebookLM
- Zoom AI Companion
Governance, accountability & review
Required for Medium/High uses
- Document a Human‑in‑the‑Loop review (Use HIL-003).
- For High‑risk scenarios involving nonpublic data, obtain pre‑approval with OIT Security & OHR.