AI for the Workplace
Date Time: 2026-09-24 11:45:47
Location: St. George, Utah
Instructor: Trevor Cahoon
Content Date: 2026-09-24
Summary
This lecture series by Trevor Cahoon offers practical, mindset-oriented guidance for using AI chatbots (like ChatGPT, Copilot, Google Gemini, and Claude) effectively in professional workplace contexts. Cahoon emphasizes that AI is a powerful supportive tool—akin to having "a thousand interns"—that augments human expertise rather than replacing it. He cautions against blindly trusting AI's first drafts, as it can be "confidently wrong," and stresses the importance of user oversight, critical judgment, and iterative refinement.
The sessions cover a range of applications, from personal tasks like generating recipes to professional uses such as drafting emails, transcribing meetings, creating action items, and analyzing large datasets like budget spreadsheets and city council minutes. A core theme is the "garbage in, garbage out" principle, with Cahoon introducing a five-part framework (Task, Who, What may use it, What must it flag, What would help) for crafting specific, context-rich prompts that yield superior results. He also discusses advanced techniques like assigning roles to the AI and using placeholders to prevent "hallucinations" (fabricated information).
A significant portion is dedicated to data privacy and security. Cahoon strongly advises against inputting any protected, sensitive, or personally identifiable information into AI models due to the risk of data breaches and misuse. He recommends using AI for public records (budgets, meeting minutes) and scrubbing any sensitive data before analysis. The overall message encourages users to experiment, learn the tool's capabilities through practice, and leverage it to enhance efficiency while maintaining control and responsibility for the final output.
Knowledge Points
1. Mindset and Role of AI
AI as a Tool, Not a Replacement: AI cannot replace human expertise and judgment; it should augment efficiency by handling small, repetitive, or complex analytical tasks.
"Thousand Interns" Analogy: AI is like having a thousand interns—potentially useful but varied in quality, requiring management and direction from the user, who acts as the expert.
First-Draft Caution: AI drafts quickly but can be "confidently wrong." Users must always review and edit outputs, never sending a first draft unedited. An example was shared where AI incorrectly inserted a "party" detail into a message for a city council.
AI as a Learning Model: AI models adapt to your usage over time. The more you use them, the better they understand your needs, style, and common tasks, sometimes even anticipating them (e.g., auto-formatting an email draft).
2. Crafting Effective AI Prompts
"Garbage In, Garbage Out" Principle: Vague, minimal prompts yield generic, useless outputs. Detailed, context-rich prompts produce significantly better results.
Five Ingredients for a Good Prompt:
Task: Clearly state what needs to be done.
Who: Define the target audience.
What may use it: Specify the output's format or medium.
What must it flag: Indicate constraints, items to avoid, or information to highlight.
What would help you: Provide additional context or information to assist the AI.
Assigning a Role: Instruct the AI to act in a specific role (e.g., "You are the training coordinator") to improve the relevance and tone of its output.
Preventing "Hallucinations": To avoid fabricated information, explicitly instruct the AI: "Please do not make information up. If you don't know it, give me a placeholder."
Iterative Refining: Use follow-up instructions to adjust tone, complexity, and format. Keeping the conversation in the same chat thread preserves context for better-tailored responses.
3. Common and Advanced AI Applications
Communication Assistance: Crafting polite emails, plain-language notices, and sensitive communications (e.g., a clock-in reminder for a new, young employee).
Administrative Support: Generating minutes and action items from rough notes or meeting transcripts, and helping structure complex schedules.
Analysis and Learning:
Comparing Procedures: Contrasting protocols to identify differences.
Learning Terms: Explaining technical terminology in accessible formats (e.g., "Explain it like I’m a five-year-old").
Data Analysis: Analyzing large datasets like budgets to identify spikes, trends, and anomalies. The AI can even hypothesize reasons for changes if given proper context.
Deep Data Analysis: Analyzing large volumes of text (e.g., meeting transcripts, survey data) to uncover underlying sentiments or themes, such as what senior leadership truly thinks about the organization.
Meeting Transcription and Analysis: Using tools to record, transcribe, and summarize meetings, automatically generating checklists of action items and drafting follow-up emails.
4. Data Privacy and Security
Prohibition on Protected Information: It is a significant security risk to input protected or personally identifiable employee information (e.g., disciplinary hearing transcripts) into an AI model.
Risk of Data Cross-Contamination: Sensitive information entered into an organizational tool could potentially be accessed by other users, leading to serious privacy breaches.
Handling Sensitive Data: If analysis is essential, all identifying information must be thoroughly "scrubbed" and redacted. The safest rule is: if the information is protected, do not use it in the AI.
Acceptable Data: Public records like budgets, city council meeting transcripts, and public meeting minutes are generally safe to use for analysis.
Organizational Policies: Even with enterprise accounts that claim data is not used for training, it's wise to remain cautious and avoid inputting any information that cannot leave the organization.
5. AI Tools and Features
AI Chatbot Landscape:
ChatGPT: Cahoon’s primary tool; adaptable and performs well across general tasks.
Claude: Strong at structured clerical tasks.
Google Gemini: Excels in research capabilities.
Microsoft Copilot: Functional but can feel like a simplified version of ChatGPT.
"Projects" or Custom AI: Features that allow users to create a contained knowledge base by uploading specific documents (e.g., city code, personnel policies), enabling the AI to answer questions based only on those sources.
The "Work" Feature: An advanced feature in some platforms (like ChatGPT) that can perform long-running, complex jobs, such as reviewing all public meeting minutes from a website to compile a list of action items.
Transcription Tools: Dedicated devices (like "Plod") and apps can record and automatically transcribe audio from meetings, which can then be fed into an AI for analysis.
6. Demonstrations and Practical Guidance
Experiment and "Play": New users are encouraged to start with fun, creative tasks (writing stories, creating recipes) to learn the tool's capabilities.
Pulmonary Embolism Prompt: A base prompt ("What is a pulmonary embolism?") yielded varied but accurate definitions across different tools. The follow-up "Explain it to me like I’m a five-year-old" produced creative, child-friendly analogies (e.g., a rock in a garden hose, a toy in a straw).
Story Transformation: AI can convert explanations into engaging stories and even rhymes, demonstrating its flexibility for communication.
Budget Analysis Exercise: An in-class assignment where attendees use AI to analyze a budget spreadsheet, reinforcing the need to provide context (e.g., new hires) for an accurate analysis.
Assignments
1. Select an AI chatbot (ChatGPT, Copilot, Google Gemini, or Claude). Run the prompt “What is a pulmonary embolism?” and compare your result with peers to observe variability. Follow up with “Explain it to me like I’m a five-year-old.”
2. Identify one repetitive work task you dislike (e.g., minutes, scheduling, polite emails). Use AI to produce a first draft, then critically review and edit it.
3. Using the provided budget spreadsheet, work in a group to analyze it with AI. Provide context (e.g., new hires) and instruct the AI to identify and explain spikes and trends, as if preparing a report for a superior.
4. Create a plain-language notice for a common internal communication (e.g., clock-in reminders for new staff) using AI, ensuring sensitivity and clarity. Verify all details before distribution.
5. Use AI to compare two procedures relevant to your department. Document the differences and validate the findings against official sources.
6. Practice follow-up prompting. First, express confusion about a topic to get a simpler explanation. Then, demonstrate expertise by asking for a targeted detail to observe how the AI shifts its tone and complexity.
