The most common use of AI in administrative work is conversational searches. A conversational search allows the user to ask questions in natural language and receive responses that are more interactive and contextual. It allows follow-up questions and maintains context from previous queries, creating a more dynamic search experience than a static internet search.

Limitations

AI conversational searches may struggle with real-time information, or rely on limited datasets compared to broader search engines. They may have overly generalized, incomplete, or biased outputs. They may also misunderstand questions. It is important not to put personal identifiable information (PII), any data protected by university, NSHE, or state or federal laws or policy into an AI tool. It is best practice to anonymize your queries (remove names, institutional data, identifying information, medical data, financial records, phone numbers, addresses, etc.)

AI can be a powerful tool for brainstorming and ideation by helping generate, refine, and organize ideas. AI-driven tools can assist in creating mind maps, suggesting creative approaches, or offering alternative perspectives on a topic. It can quickly analyze large volumes of information, highlight relevant trends or insights, and propose new avenues for exploration. This can be especially useful for collaborative projects, research development, or curriculum design. AI can support rapid iteration and enhance the diversity of ideas.

Limitations

AI models can sometimes perpetuate biases based on the data they are trained on which can lead to skewed, unrepresentative, or incomplete ideas. It can also overlook novel or unconventional approaches. Models can miss nuance and subtleties of the human experience, or cultural considerations.

AI-powered tools can auto-suggest code completions, convert descriptions in plain language into functional code snippets, suggest refactoring changes, find and fix bugs, convert between coding languages, create documentation and code explanations, automate repetitive tasks (data processing, file handling, automation scripts), and more.

Limitations

Always check the code quality and accuracy, and ensure there are no security vulnerabilities. Be cautious entering code with sensitive data or proprietary code. Check the AI privacy policies to ensure the tool does not store or use the data to improve its model. Understand licensing implications and be mindful of attributions. Also, make sure the code is designed for accessibility, it is unbiased, and you are transparent about the use of AI in your code. Remember, the responsibility for code is with the developer/programmer.

AI-powered tools can prioritize emails, suggest replies, automate responses to common inquiries. They can also help extract entry of data from forms, documents, databases, and spreadsheets. AI can also help generate standard documents based on templates. Content summarization is another popular administrative use of AI, by providing content summaries. At UNLV, Zoom’s AI Companion tool can be used for meeting summaries.

AI can also be applied to workflows to automate processes.
UNLV’s Scarlet chatbot incorporates AI as a virtual assistant to handle inquiries from students, faculty, and staff, and guide them to resources and information.

Limitations

Ensure that any data input into AI models follows any policies set by NSHE, UNLV, or any state or federal laws. Ensure outputs are free from bias, offensive or malicious content, and ensure all outputs are vetted for applicability to the university setting. Use is not recommended for academic decision-making, evaluations, admissions, or other high stakes use cases unless users can be sure the AI system is secure, free from biases, and compliant for data privacy. UNLV does not currently have an enterprise AI system. Users are responsible for AI output. For example, always read meeting summaries before sending them out and edit for privacy, appropriateness, and tone. Work with your department, unit, or college for guidelines on AI use specific to them.

AI can generate complex Excel formulas based on natural language descriptions (e.g., “Find the total expenses for each department over the last year.”, analyze formulas for errors and suggest corrections, remove duplicates, fill in missing values, or standardize data formats.  It can also apply conditional formatting rules, summarize large data sets, optimize pivot table formulas, and combine functions to automate workflows.

Limitations

AI may struggle with a contextual understand of the data, that may lead to suggestions that are not relevant to the data or administrative need. It may not generate the most efficient formulas for highly complex data analysis or models, and may struggle with nuanced Excel features like nested functions, dynamic arrays or macro integration. Of course, results depend of clean, well-organized data. Utility may be limited if AI is not integrated into other systems.

AI can analyze data and provide recommendations based on trends, patterns, and predictive models. machine learning and natural language processing, AI can identify patterns and trends within large datasets. AI can help administrators make more informed decisions and improve institutional efficiency.

Limitations

Large datasets may include sensitive information that requires sanitization (FERPA, HIPPA or other PII). AI models can also perpetuate biases, especially of data is also incomplete or biased. Many AI algorithms are not transparent about how decisions are made, which can be problematic for accountability and trust. The human element and accountability for final outcomes and decisions is essential. Implementing and maintaining AI solutions may also not be sustainable in terms of cost or resource requirements.

AI tools can be valuable in creating documentation by automating or augmenting tasks. AI can pull data from datasets to generate reports, extract data from forms, track versions and changes, suggest updates, and even ensure documentation complies with existing standards. It can also be used to generate templates, automate responses in chatbots, and assist with building knowledge base articles.

Limitations

AI-generated documentation needs to be checked for errors or inaccuracies, particularly with complex or high-impact information. AI may struggle to fully adapt to UNLV-specific needs or formats without significant human intervention. As always, data privacy and security and the handling of sensitive information must be considered, as well as compliance with privacy and security policies, laws, and guidelines from the institutional, local, state, or federal level.

AI image generation can be useful for promotional materials, generating alt text descriptions for accessibility, visualizing data, analyzing images, facility planning and maps, photo editing, training and onboarding materials, and other applications.

Limitations

AI models can perpetuate biases by outputting biased or inaccurate images. Images generated by AI may also lack or alter important details, or generate images with incorrect or inconsistent elements. Images may also raise issues with intellectual property (IP), copyright, and usage rights.

As we continue to explore the potential of AI, UNLV IT is committed to ethical usage that respects privacy, promotes equity, and aligns with UNLV mission and values.