GenAI Guidelines for Staff

• Automation of routine tasks: Identify administrative tasks that are repetitive and time-consuming, such as cleaning data, organizing, formatting and report generation.
• Helping with communication: Utilize GenAI for drafting emails, drafting text content for newsletters or editing speaker notes for presentations.
• Data gathering and reporting: Leverage GenAI for collecting and cleaning data, or making a first pass at reporting. Always follow U-M’s policy on data stewardship.
• Exploration: Use GenAI tools as a thought partner to kickstart creativity or strategy to potentially spark an idea for the user to then build upon.
• References: For transparency, cite and explain your use of GenAI tools in all work products. Verify that the work was spot-checked for accuracy and fairness.

• You are responsible for your work products. Think of AI as a teammate that complements your expertise and amplifies your skills, rather than a substitute for it.
• Bias and discrimination. GenAI systems may exhibit or amplify biases, cultural misrepresentation, mis/disinformation–leading to discriminatory results. Learn more about this and ways to mitigate bias through U-M resources and feedback mechanisms.
• Critical decisions and analysis without human oversight: GenAI should not make critical decisions without human analysis and continued engagement.
• Remain vigilant regarding over-reliance on GenAI: For tasks requiring personal judgment, empathy and nuanced understanding, it is best practice to rely on humans as the final decision-makers.
• Using unsanctioned GenAI tools and sensitive data: It is U-M policy to use University-approved tools with sensitive data. Please review the ‘Ensuring Data Privacy’ section of these guidelines for more information.

• We’re beginning to see AI integrations into U-M’s enterprise applications, like Zoom’s AI Companion. That integration was facilitated by ITS, which has a process for vetting emerging AI features for all of U-M’s enterprise applications like Adobe, Slack and Google Drive.
• When considering existing or emerging AI features for the unique applications that your team uses, please consult the ITS AI Service Team for advice on GenAI tools.

• Consider the processes that enable your work and the stakeholders involved. Examples could include:
  • Meeting scheduling, note-taking and delegation of action items.
  • Collecting and evaluating data.
  • Reviewing and updating website content.
  • Reviewing research proposals or grant applications
  • Drafting emails or templatized documents.
• Compare how the steps of these processes align with the ‘Appropriate Uses’ guidelines outlined in this document.

• Changing or updating an existing system should be done with great care and leadership support.
• Run pilots, measure results and discuss outcomes.
• Other teams across the University and at peer institutions may be exploring or have already made integrations that seem appealing. Investing in benchmarking may save you from reinventing a proven integration.
• Sharing success stories can help show others what is possible and enable a cognitive onramp for adoption.
• Increasing individual and team GenAI literacy will help when navigating use cases and integrations.

• Overview of what GenAI is.
• Overview of U-M’s GenAI tools.
• Privacy risks and how U-M’s GenAI tools protect data.
• Ethical considerations.
• Prompting best practices.
• How-to spot-check outputs for accuracy, appropriateness and bias.

Also, explore ITS AI Services Support and these workshops from ITS.

How GenAI applies to specific job functions is a more nuanced learning journey. Explore examples, scenarios, conferences and case studies that apply to your role.

• Regardless of job function, be aware that copyright and plagiarism concerns in this space are not well defined.
• Request a consultation from the ITS Service Team.
• Join the U-M AI Community of Practice and Slack channel to network and learn from colleagues.

• When using GenAI tools, handle sensitive data with care, limiting its use to only what is necessary for your task.
• Review the Safe Computing site to confirm what types of data are acceptable to use with U-M’s AI tools.
• Responsibility for data privacy falls on individual users; be aware of U-M’s policies and seek guidance if needed.
• Exercise caution and transparency with GenAI features in U-M enterprise software like Zoom, Adobe and Google Drive.
  • Alert team members when AI tools are being used in meetings and projects.
• When considering existing or emerging AI features for the unique applications that your team uses, please consult the ITS AI Service Team for advice on GenAI tools.

• Individual Responsibility of Work: The person responsible for the work product needs to manually check for quality and accuracy.
  • Work submitted with the use of AI may reflect back on individual and institutional reputation.
  • AI systems that may have a history of being accurate and effective, can at any time provide information that is no longer fair or useful.
    • Human spot-checks must be executed with every use.
• Individual Choice and Adoption: Employees may have concerns about bias, sustainability, quality, etc and are not required to use AI in their work.
  • Staff who choose to use AI tools should thoughtfully explore the risks and opportunities of using GenAI in their work.
• Accessibility: Employees should have access to AI tools and training relevant to their work. All of U-M’s GenAI tools and training should adhere to SPGs around accessibility.
• Surveillance: Due to concerns about false positives, the use of GenAI detection software is not advised.
• Errors and Irresponsible Use: People will make mistakes and need guidance, and that should feel like a positive learning experience. When errors occur, seek to understand what mix of human and machine led to the issue. Take appropriate action to fix the error and adjust workflows to avoid making similar mistakes in the future.
  • Managers should seek to support employees who are struggling to get up to speed or need accommodations.