Welcome

Large Language Models and Gen Z: A Match Made in Heaven?

What You'll Learn

This guide teaches one core idea: giving AI the right context transforms it from a slop machine into a genuine collaborator. Along the way, you'll discover:

What Gen Z actually thinks about feedback and AI (it's not what you'd expect)
The Mentor Mindset framework for high-standards, high-support leadership
Why context is the difference between thoughtful and lazy AI use
A three-step method: Context → Prompt → Output
Practical demos with NotebookLM, Claude, ChatGPT, and other platforms
Real-world use cases combining AI with mentorship frameworks

Key Principle: Context Changes Everything

When AI gives you a generic, sloppy response, the problem usually isn't the AI—it's that you expected it to read your mind. The shift from "AI as oracle" to "AI as informed collaborator" starts with providing real context: your audience, your goals, your constraints. This guide shows you how, with Gen Z communication as the running case study throughout.

Chapter 1 — Understanding Your Trainees

The Gen Z Perspective

Understanding how Gen Z radiology trainees think about feedback and AI is essential for effective communication. This generation brings unique attitudes shaped by growing up with technology—but that doesn't mean they automatically embrace AI.

Who is Gen Z? Born between 1997 and 2012, Gen Z are your current medical students, interns, and junior residents. They're the first generation to grow up entirely in the smartphone era, with social media and instant information access as constants in their lives.

How Gen Z Wants Feedback

Gen Z seeks trust, but it must be earned through consistent actions. They value transparency, authenticity (genuine care), and visible human effort behind outreach. Generic mass communications are quickly spotted and dismissed. They want to understand the "why" behind decisions and expect two-way dialogue rather than top-down criticism.

How Gen Z Views AI Use

Contrary to assumptions, Gen Z is actually skeptical and anxious about AI. They've grown up seeing AI's limitations firsthand and are wary of "AI slop"—generic, low-effort AI-generated content that lacks authenticity. They accept AI for low-level administrative tasks but strongly favor when AI augments rather than replaces human skills and judgment.

Continue Teaching Human Skills

Despite AI's capabilities, Gen Z still needs and values instruction in foundational human skills: clinical reasoning, patient communication, procedural techniques, and professional judgment. Position AI as a tool that enhances these skills, not a shortcut that makes learning them obsolete.

Key Attitudes and Best Practices for Radiology Faculty +

Understanding Gen Z's Mindset

Skeptical but Open: Gen Z doesn't automatically trust technology or authority. Trust must be earned through consistent, authentic interactions. Don't assume they'll be impressed by AI-generated content simply because it's AI.

AI for Low-Stakes Tasks: They're comfortable with AI handling administrative work, scheduling, and routine communications—but not replacing human judgment on clinical matters or important feedback conversations.

Authenticity is Non-Negotiable: They can spot inauthenticity quickly. Any communication—AI-assisted or not—needs to feel genuine and caring. Personalization matters more than polish.

They Value Earned Respect: Gen Z trainees are looking for recognition when they meet high standards. Acknowledgment of their growth and achievements builds trust and motivation.

Best Practices for Radiology Educators

Do This	Avoid This
Be transparent about when you use AI assistance	Pretending AI-assisted content is fully human-written
Keep human effort visible in feedback and outreach	Sending obvious mass AI-generated messages to all residents
Model AI as augmenting clinical reasoning skills	Positioning AI as a replacement for learning fundamentals
Continue teaching foundational radiology interpretation	Assuming AI makes traditional reading skills obsolete
Provide specific, actionable feedback with context	Generic feedback that could apply to anyone
Recognize achievements and growth explicitly	Only giving feedback when there are problems

Key Takeaway

When working with Gen Z radiology trainees, position AI as a tool that enhances your ability to connect authentically—not as a shortcut that replaces genuine human engagement. They can tell the difference.

Chapter 2 — A Framework for Feedback

The Mentor Mindset

David Yeager's research identifies three distinct mindsets educators adopt when giving feedback. Understanding these mindsets helps you communicate more effectively with Gen Z trainees—whether you're using AI assistance or not.

About This Framework: David Yeager is a developmental psychologist whose research focuses on how to motivate young people effectively. His "Mentor Mindset" framework comes from studying what makes feedback actually land with the next generation.

Mentor Mindset

High Standards, High Support

Combines rigorous expectations with genuine belief in the trainee's ability to meet them. Communicates both the challenge and the confidence. This is the most effective approach for Gen Z.

Enforcer Mindset

High Standards, Low Support

Feedback feels like an attack or humiliation. Destroys the trainee's sense of respect and dignity. May produce short-term compliance but damages the relationship and long-term development.

Protector Mindset

Low Standards, High Support

Feedback lacks challenge to grow. Implies low confidence in the trainee's capability. Feels patronizing and ultimately undermines their professional development.

"I am giving you this feedback because I hold you to a high standard, and I'm confident you can meet it—and I want to help you get there."

Applying the Mentor Mindset in Radiology Education +

The Two-Part Formula

The Mentor Mindset works by communicating two messages simultaneously:

Part 1 — High Standards: "I expect excellent work from you because that's what our patients deserve and what you're capable of." Be clear about what excellence looks like in radiology.

Part 2 — High Support: "I believe you can meet these standards, and I'm here to help you get there." Offer specific guidance, resources, and availability.

Practical Examples for Radiology

Scenario: Resident missed a finding on a study

Mindset	What It Sounds Like
Enforcer	"How did you miss this? This is basic stuff. You need to be more careful."
Protector	"Don't worry about it, these things happen. It was a subtle finding."
Mentor	"Let me show you what I saw here. This is a challenging finding, but it's something you need to catch consistently. Let's work through the search pattern that would help you identify this in the future. I know you can get this."

Giving Back Earned Respect

A crucial part of the Mentor Mindset that many educators overlook: when trainees meet your high standards, explicitly acknowledge it. Gen Z trainees are looking for earned respect from their mentors—recognition that they've risen to the challenge.

Example acknowledgment: "I noticed your reports have gotten much more precise over the past month. Your differential diagnoses are more focused and your clinical correlations are spot-on. That's exactly the growth I was hoping to see. Keep it up."

Why This Framework Isn't Perfect—But Helps

No single framework captures every communication challenge with Gen Z. Individual trainees vary significantly, and context matters. However, the Mentor Mindset provides a useful default that aligns with what Gen Z values: high expectations combined with genuine belief in their potential and concrete support.

Mentor Mindset Quick Reference

Formula: High Standards + High Support + Genuine Belief
Key phrase: "I'm giving you this feedback because I believe you can meet a high standard"
Don't forget: Give earned respect when standards are met
Avoid: Enforcer (harsh, unsupportive) or Protector (soft, low expectations)

Chapter 3 — The Superpowers

What AI Does Exceptionally Well

Despite valid concerns about AI, it has genuine superpowers that can help bridge communication gaps—including with Gen Z trainees. The key is knowing what AI excels at and using it strategically.

Tone and Empathy

AI excels at adapting communication style to different audiences. It can help you strike the right balance between professional and approachable, authoritative and supportive—essential for the Mentor Mindset.

Judgment and Perspective

AI can provide perspective on decisions and tradeoffs by considering multiple angles simultaneously. It surfaces considerations you might miss when you're focused on immediate clinical demands.

Reorganizing Information

AI is excellent at restructuring information for clarity and impact. It can take disorganized thoughts and transform them into coherent narratives—useful for feedback, presentations, and documentation.

Understanding Nuance

Modern AI handles nuance, context, and ambiguity remarkably well. It can read between the lines and navigate complex interpersonal situations when given adequate context.

Remember

These superpowers only work if you give AI what it needs. Generic prompts produce generic results. The next chapters show you how to provide the context that makes AI genuinely useful.

Chapter 4 — The Fix

Context is Everything

Context is the difference between thoughtful and lazy AI use. When you get a generic, sloppy AI response—what went wrong? Usually, you expected AI to read your mind. It can't. But give it proper context, and everything changes.

The Big Shift: Modern AI can now handle very large amounts of information (called "tokens"). This means you can give it extensive context—entire documents, multiple files, detailed background—and it will use all of it to give you better responses.

What Are Tokens?

Tokens are the units AI uses to process text—roughly equivalent to words or word pieces. Early AI could only handle a few thousand tokens (a few pages). Modern AI can handle 100,000+ tokens (entire textbooks worth of content). This means you can now upload entire rotation evaluation forms, curriculum documents, or detailed case histories and AI will use all of it.

The Mind-Reading Problem

When AI gives you bad output, ask yourself: "Did I expect AI to know things I never told it?" Generic prompts produce generic results. Specific context produces specific, useful results. AI can't know your department culture, this resident's history, or your communication style unless you tell it.

From Oracle to Collaborator

Stop thinking of AI as a magic oracle that should just "know" what you need. Instead, treat it as an informed collaborator—like briefing a new colleague. The more you share about your situation, goals, and constraints, the better it can help.

Context Checklist for Radiology Educators

Who is the trainee? (Level, background, strengths)
What are you trying to accomplish? (Feedback, guidance, difficult conversation)
What's the situation? (Rotation, specific incidents, history)
What constraints exist? (Program policies, documentation requirements)
What framework do you want to use? (Mentor Mindset, specific competencies)
What tone is appropriate? (Formal/informal, written/verbal)

Chapter 5 — The Method

Three Steps to Thoughtful AI Use

Here's a systematic approach to working with AI that consistently produces better results. Master these three steps and you'll avoid the "AI slop" that Gen Z trainees can spot instantly.

01

Context

Feed AI the materials it needs—research, guidelines, examples. Brief it like a new colleague, not a search engine.

+

Brief AI Like a New Colleague

Jeremy Utley's core insight is that most people treat AI like a search engine—type a short question, hope for a good answer. The people who get exceptional results treat AI like a new colleague on their first day. You wouldn't ask a new team member to write a remediation plan without first handing them the resident's evaluations, the program's policies, and your notes from the last feedback conversation. AI is no different.

This means uploading actual materials—not summarizing them, not describing them, but giving AI the real documents to work from. Modern AI tools can read PDFs, spreadsheets, and long documents directly. The more relevant material you provide, the more specific and useful the output becomes.

What to Upload

Think of the materials you'd hand a trusted colleague before asking for their help. For radiology education, that might include:

Research & Frameworks

Research articles: Upload the Gen Z research briefing, the Mentor Mindset paper, or any article relevant to the problem you're working on. AI can synthesize across multiple sources in seconds.
Frameworks: The Wise Feedback formula, the Three Lenses model, ACGME milestones—upload the actual documents so AI can apply them accurately, not from memory.
Guidelines & policies: Your program's remediation policies, rotation evaluation criteria, competency standards, institutional guidelines.

Examples & Templates

Good examples: Upload a feedback email you wrote that landed well, a remediation plan that worked, or a recommendation letter you're proud of. Tell AI: "Use this as a model."
Templates & forms: Evaluation forms, milestone tracking sheets, rotation assessment templates—give AI the actual structure to work within.
Prior work: Past evaluations, meeting notes, documented observations about a specific trainee.

Utley's Rule of Thumb

For higher quality output, never prompt with less than 400 words of context—including your objectives, audience, examples, and constraints. If your prompt is shorter than a few sentences, you're probably expecting AI to read your mind. Upload the materials, state your goal, and let AI do what it does best: synthesize, reframe, and draft from real sources.

Source: Jeremy Utley — AI as Teammate, Not Tool (YouTube)

02

Prompt

You don't need to master prompting—just learn three levels and let AI help you write the rest.

+

Three Ways to Prompt

Start quick, add structure only when you need it. Most prompts only need the first level.

Quick (good enough for most tasks)

Persona — who the AI is ("You are a radiology educator…")
Task — what you want produced ("Write mid-rotation feedback…")
Context — audience, constraints, background ("PGY-2 resident, strong on image interpretation, struggles with differential diagnosis")

Structured (when you need precision — add these)

Exemplar: Paste a sample of what good looks like ("Copy this style")
Format: Bullets vs. paragraphs, length, sections, output type
Tone: Formal/casual, concise/detailed, friendly/authoritative

Full template (copy-paste and fill in): Persona: [who the AI is] Task: [what you want produced] Context: [audience, constraints, background] Exemplar: [a sample to imitate] Format: [shape of the output] Tone: [voice and style]

Let AI Write the Prompt for You

Instead of crafting the perfect prompt yourself, describe what you want in plain language and ask the AI to write a better prompt for you. AI is better at writing structured prompts than most humans—it will add context, constraints, and formatting you didn't think to include.

Prompt generator: "I'm trying to get good results using the following prompt: [paste your prompt]. Your task is to write a better prompt that would produce better results. Give me 3 variations."

Or just describe what you need: "I need to prepare for a mid-rotation feedback session with a resident who has both strengths and areas for improvement. Help me write a detailed prompt that will get me the best possible feedback outline and talking points. Ask me any clarifying questions first."

AI will ask you targeted questions about the trainee, the rotation, specific observations, and your goals—then generate a comprehensive prompt you can use, often better than what you'd write yourself.

Use XML Tags for Complex Prompts

XML tags act like labeled folders that help AI know exactly which part of your prompt is which. Instead of one wall of text, you separate your instructions into clear sections. You don't need this for simple questions—use it when your prompt has multiple inputs or when you keep getting muddled output.

Before (everything mashed together): You are a medical educator. Here is the learner's case presentation. I want you to give feedback using the Wise Feedback method. Be direct but supportive. The learner is a PGY-2...

After (tagged and organized): <persona>You are an experienced radiology educator.</persona> <context>The learner is a PGY-2 resident presenting their first complex case to the team.</context> <task>Give feedback on the case presentation below using the Wise Feedback method: state the standard, affirm capability, offer a path forward.</task> <tone>Direct but supportive. High standards, high respect.</tone> <case_presentation> [paste the learner's presentation here] </case_presentation>

You can name the tags anything—<audience>, <constraints>, <data>—as long as you're consistent. Think of it as giving AI a filing system instead of a pile of papers.

Fix Bad Output

When AI gives you something weak, don't start over. Use one of these patterns to steer it:

Three Debugging Patterns

Chain of Thought: "Think step by step, show your reasoning, then give the final answer."
Verifier: "Ask me 3 clarifying questions before answering."
Refinement: "Propose 2 sharper versions of my question. I'll pick one."

Check the Output (OCEAN)

Before you use AI output, run it through these five checks:

OCEAN Framework

Original — Are the ideas non-obvious, not generic?
Concrete — Are there real names, numbers, examples?
Evident — Is the reasoning visible and traceable?
Assertive — Does it take a stance, not hedge everything?
Narrative — Does it flow logically from start to finish?

Two Meta-Tips

1. Don't write the prompt yourself — describe what you want and ask AI to write the prompt for you. It's faster and usually stronger.

2. Use XML tags for complex prompts — simple questions don't need structure, but multi-part requests benefit from labeled sections.

Sources: Jeff Su — Master AI Prompting (YouTube) · MIT Monk — AI Prompting Techniques (YouTube)

03

Output

Expand beyond simple text! AI can produce scripts, infographics, audio summaries, slide decks, and much more.

+

Most people ask AI to write an email or a paragraph and stop there. But the same context you already provided can produce dramatically different—and often more useful—formats. Ask yourself: what form would actually be most helpful here?

Creative Output Types

Output Type	What AI Produces	Example in Radiology Education
🎬 Scripts	Word-for-word dialogue with stage directions, speaker labels, and timing cues	"Write a 5-minute script for a difficult feedback conversation using the Wise Feedback method, with lines for both attending and resident"
📊 Infographics	Visual layouts with sections, icons, and data points (as code you can render or hand to a designer)	"Create an infographic summarizing the Mentor Mindset framework—four quadrants, two types of respect, and the Wise Feedback formula"
📑 Slide Decks	Complete slide outlines with speaker notes, or full presentations you can export	"Build a 10-slide deck for a faculty development talk on Gen Z communication preferences, with speaker notes for each slide"
🎧 Audio Summaries	Podcast-style scripts or audio briefings (NotebookLM can generate these directly)	"Generate an audio overview of our department's new AI policy that I can share with residents as a 5-minute podcast-style briefing"
📋 Checklists & Rubrics	Structured evaluation tools with criteria, scoring levels, and observable behaviors	"Create a feedback checklist I can use during readout sessions—observable behaviors mapped to ACGME milestones"
💬 Conversation Simulations	Interactive role-play where AI plays the trainee so you can rehearse	"Act as a defensive PGY-2 resident. I'll practice delivering remediation feedback. Push back realistically so I can refine my approach"
📧 Multi-Format Campaigns	The same content adapted for different channels in one request	"Take this policy update and give me: a 2-sentence Slack message, a formal email to faculty, and 3 talking points for the next department meeting"

Try this — same context, three different outputs: You've uploaded Gen Z research and a resident's evaluations. Now ask for: 1. "Write a script for a 10-minute feedback conversation" 2. "Create a one-page visual summary I can print and reference" 3. "Generate a self-assessment questionnaire I can give the resident before we meet" Same materials, three completely different—and complementary—deliverables.

Think Format First

Before you type your prompt, ask: "What format would actually be most useful?" A script you can rehearse from is often more valuable than a paragraph you'll read once. An infographic residents can pin to their workstation outlasts any email. A conversation simulation that lets you practice beats reading advice about how to handle a difficult talk.

Explainer: AI Chatbots vs. AI Agents — What's the Difference? +

Understanding the Distinction

As AI tools evolve, you'll hear the terms "chatbot" and "agent" used differently. Understanding the distinction helps you choose the right tool for the job.

AI Chatbots (Interactive Assistants)

Chatbots are conversational AI tools that respond to your questions and requests in real-time. You have a back-and-forth dialogue, refining your request and getting immediate responses.

How it works: You ask → AI responds → You clarify → AI refines → Repeat
Best for: Brainstorming, drafting, getting feedback on ideas, learning
Examples: ChatGPT, Claude, Gemini in conversation mode
Radiology use: "Help me draft feedback for this resident" with back-and-forth refinement

AI Agents (Autonomous Workers)

Agents are AI systems that can execute multi-step tasks with minimal supervision. You give them a goal and context, and they work autonomously to accomplish it—making decisions, using tools, and producing outputs.

How it works: You define goal + context → AI plans steps → AI executes autonomously → Delivers result
Best for: Complex, multi-step workflows; research tasks; systematic analysis
Examples: Custom GPTs, Claude Projects with instructions, Gemini Gems
Radiology use: "Analyze this resident's last 6 months of evaluations and create a development plan" where the AI reads multiple documents and synthesizes a comprehensive output

When to Use Each

Situation	Use This
You want to explore ideas and aren't sure what you need yet	Chatbot
You have a clear, well-defined task with multiple steps	Agent
You need real-time collaboration and refinement	Chatbot
You want to hand off a task and come back to a finished product	Agent
You're learning how to use AI effectively	Chatbot (start here)

Practical Advice

Start with chatbots to build your skills, then graduate to agents as you get better at providing context. Agents are powerful but require clear instructions upfront. If you can't describe what you want clearly, use a chatbot conversation first to figure it out.

Chapter 6 — Tools and How-To Guides

AI Platforms and Tools

Different AI platforms have different strengths. Here's a detailed guide to the major options, step-by-step instructions for getting started, and guidance on when to choose which.

Google NotebookLM

Best for learning context-first AI — Free

▼

What it is: NotebookLM is Google's AI research assistant that only uses sources you provide. It doesn't search the web—forcing you to practice context-first AI use. This makes it ideal for learning good habits.

Why It's Great for Radiology Educators

Because NotebookLM can't access external information, you must upload your context. This builds the habits this guide teaches. It's also excellent for analyzing multiple documents together—like comparing resident evaluations over time or synthesizing curriculum materials.

Key Features

Source-grounded responses: Every answer cites which source it came from
Multiple source types: Upload PDFs, Google Docs, websites, YouTube videos
Audio Overview: Creates podcast-style summaries of your sources
Notebook organization: Keep different projects (rotations, residents) separate

Step-by-Step: Getting Started

Go to notebooklm.google.com
Sign in with your Google account (personal or institutional)
Click "New Notebook" to create your first workspace
Add sources by clicking the "+" button:
- Upload PDFs (rotation syllabi, evaluation forms, research papers)
- Paste Google Docs links
- Add website URLs
- Include YouTube video links
Start chatting in the text box — ask questions about your sources
Click citations to see exactly where information came from

Radiology Educator Use Cases

Upload all evaluations for a resident and ask "What patterns do you see?"
Add curriculum documents and ask "How does this resident's progress compare to expected milestones?"
Upload Gen Z research and ask "How should I approach feedback with this generation?"
Create podcast-style summaries of reading materials for residents

NotebookLM Tutorial Audio Overview Demo

ChatGPT

Most versatile — Free tier available, Plus $20/month

▼

What it is: OpenAI's flagship AI assistant. The most widely used AI chatbot, constantly improving, with both free and paid versions. Known for versatility and the GPT Store ecosystem.

Key Features

Custom GPTs: Pre-configured AI assistants with specific personalities and instructions
File uploads: Analyze PDFs, images, spreadsheets, and documents
Web browsing: Search the internet for current information
Image generation: Create images with DALL-E
Code interpreter: Execute Python for data analysis
Voice mode: Have spoken conversations with AI

Step-by-Step: Getting Started

Go to chat.openai.com
Create an account (free) with email, Google, or Microsoft
Start a new chat by typing in the message box
Upload files by clicking the paperclip icon next to the text box
Explore GPTs by clicking "Explore GPTs" in the sidebar

How to Create a Custom GPT (ChatGPT Plus)

Click your profile → "My GPTs" → "Create a GPT"
Describe what you want: "I want a GPT that helps me write resident feedback using the Mentor Mindset framework"
Upload relevant documents (evaluation forms, frameworks)
Test and refine through conversation
Save and use whenever you need it

Radiology Educator Use Cases

Create a "Resident Feedback Writer" GPT with your program's templates
Upload case images for teaching material generation
Use Code Interpreter to analyze resident performance data
Draft recommendation letters with consistent structure

Beginner Guide Custom GPTs Tutorial

Claude

Best for long documents and nuanced writing — Free tier, Pro $20/month

▼

What it is: Anthropic's AI assistant, known for nuanced responses, strong writing ability, and exceptional handling of long documents.

Key Features

Projects: Organize work with persistent context across conversations
Massive context window: Analyze very long documents (200K+ tokens)
Artifacts: Interactive documents, code, and visualizations
Strong writing: Nuanced, professional content with excellent tone control
Analysis strength: Particularly good at synthesizing complex information

Step-by-Step: Getting Started

Go to claude.ai
Create an account with email or Google
Start chatting in the message box
Upload files by clicking the attachment icon or dragging files

How to Use Claude Projects (Claude Pro)

Click "Projects" in the left sidebar
Click "New Project" and give it a name
Add project instructions: Describe the context Claude should always have
Upload project knowledge: Add documents that apply to all conversations
Start conversations within the project — context persists across all chats

Radiology Educator Use Cases

Create a Project for each rotation with relevant syllabi pre-loaded
Upload entire curriculum documents for comprehensive analysis
Use for writing nuanced, sensitive feedback
Analyze multiple years of resident evaluations at once

Claude Basics Projects Tutorial

Google Gemini

Best for Google Workspace integration — Free tier, Advanced $20/month

▼

What it is: Google's AI assistant, deeply integrated with Gmail, Docs, Sheets, and other Google services.

Key Features

Gems: Custom AI personas with specific expertise
Google integration: Access Gmail, Drive, Calendar directly from chat
Image understanding: Strong multimodal capabilities
Extensions: Connect to YouTube, Maps, Flights, and more
Deep Research: Comprehensive research reports with citations

Step-by-Step: Getting Started

Go to gemini.google.com
Sign in with your Google account
Enable extensions by clicking the puzzle piece icon
Start chatting — use "@Gmail" or "@Drive" to reference Google content

How to Create Gems (Gemini Advanced)

Click "Gem manager" in the sidebar
Click "New gem"
Give it a name and description
Write detailed instructions about how this Gem should behave
Add example interactions if helpful
Save and access from your Gem library

Radiology Educator Use Cases

Create a "Feedback Helper" Gem with the Mentor Mindset framework
Use @Drive to access evaluation documents in Google Drive
Ask about YouTube educational videos for resident recommendations
Summarize emails from residents or program coordinators

Complete Guide Creating Gems

Microsoft Copilot

Best for Microsoft 365 integration — Free tier, full integration requires M365

▼

What it is: Microsoft's AI assistant, integrated into Windows, Edge, and Microsoft 365 apps.

Key Features

Office integration: AI assistance directly in Word, Excel, PowerPoint, Outlook
Email drafting: Compose, summarize, and respond to emails in Outlook
Meeting summaries: Automatic recaps of Teams meetings
Document assistance: Write, edit, and format in Word with AI help
Enterprise security: Data stays within your organization's boundary

Step-by-Step: Getting Started

Free version: Go to copilot.microsoft.com or use the Copilot sidebar in Edge
In Office apps: Look for the Copilot icon in the ribbon (requires M365 Copilot license)
In Word: Click Copilot → "Draft with Copilot" to generate content
In Outlook: Click Copilot when composing to get draft suggestions
In Teams: Use Copilot during or after meetings for summaries

Using Copilot in Word for Feedback

Open a new Word document
Click the Copilot icon in the ribbon
Select "Draft with Copilot"
Describe what you want: "Write mid-rotation feedback for a PGY-2 radiology resident on chest imaging rotation..."
Review, edit, and refine the output

Radiology Educator Use Cases

Draft evaluation narratives directly in Word
Summarize long email threads about resident performance
Generate meeting agendas and capture action items from CCC meetings
Create PowerPoint slides for educational presentations

Copilot Overview Copilot in Office

Quick Comparison: Which Platform to Choose? +

If you need...	Use...
To learn context-first AI (start here!)	NotebookLM
Research grounded only in sources you provide	NotebookLM
General-purpose assistance with web access	ChatGPT or Gemini
Analysis of very long documents	Claude
Nuanced, sensitive writing	Claude
Integration with Google Workspace	Gemini
Integration with Microsoft 365	Copilot
Custom AI personas for repeated tasks	ChatGPT (GPTs), Gemini (Gems), or Claude (Projects)
Image generation for teaching materials	ChatGPT (DALL-E)
Enterprise security requirements	Copilot (with M365 license)

Recommendation for Beginners

Start with NotebookLM to build good context-providing habits. Once comfortable, expand to ChatGPT or Claude for broader capabilities. Use platform-specific tools (Gemini, Copilot) when you need their integrations.

Chapter 7 — Applications

Real-World Use Cases

Four progressive levels for applying AI in radiology education—from basic chatbot interactions to advanced automated workflows. Each level builds on the previous one, showing how to effectively use context and frameworks.

Level 01

Chatbot + Upload Guidelines

Adding Context to AI Chatbot

Case: Bridging Generational Gaps with Gen Z Trainees using Mentor Mindset guidelines

+

The Approach: Upload research frameworks and guidelines directly to your AI chatbot (Claude, ChatGPT, Gemini) to provide context for your specific question. This is the foundational level—learning to give AI the context it needs rather than expecting it to "just know."

Example prompt: "Based on the Mentor Mindset framework and Gen Z research I've uploaded, help me rewrite this feedback for a PGY-2 resident who's struggling with overnight call efficiency. I want to be direct but also show that I genuinely care about their development and believe they can improve. They value authenticity—help me avoid sounding corporate or condescending."

How to do this

In Claude or ChatGPT: Upload PDF documents about the Mentor Mindset framework and Gen Z communication research directly in the chat. Then ask your question.

Key skill: Learning to provide relevant context documents rather than expecting AI to have this specialized knowledge built in.

Level 02

Agents Part I

Using Research Articles Directly

Case: Google NotebookLM with Mentor Mindset research articles. Gems adds web search to closed sources of NotebookLM

+

The Approach: Use specialized AI tools that work with your uploaded research as a dedicated knowledge base. NotebookLM excels at working exclusively with your uploaded sources, while Gems can combine those sources with web search for current information.

NotebookLM Strengths

Source Fidelity: Works only with your uploaded documents—no hallucination from outside sources
Citation: Every answer includes citations showing which source it came from
Multi-Document Analysis: Upload multiple research articles and it synthesizes across all of them
Audio Overview: Can generate a podcast-style discussion of your sources

Google Gems Enhancement

Web Search Integration: Gems can search the web and combine findings with your NotebookLM sources
Current Information: Get up-to-date statistics, recent studies, or current best practices
Persistent Instructions: Create a Gem with standing instructions (e.g., "Always use Mentor Mindset principles when giving feedback advice")

Example workflow: "In NotebookLM, I've uploaded three sources: the Mentor Mindset research paper, the Gen Z workplace communication study, and our program's evaluation framework. Based on these three sources, help me design a mid-rotation feedback template that incorporates all three frameworks. Then in Gems, use Chatbox which has access to web search, for responses."

When to use this level

Use NotebookLM when you need answers that come exclusively from specific research articles or documents—perfect for evidence-based approaches. Add Gems when you need to supplement your closed sources with current web information.

Level 03

AI Agent Part II

Agents Bringing Different Personas

Career Counseling: The Three Lenses Framework

+

The Approach: Advanced AI agents that can adopt multiple perspectives or personas within a single conversation. This lets you get advice from different viewpoints—like having multiple consultants analyze the same problem. Below is an example from Dylan Davis "Don't Prompt AI. Let It Prompt You." Here he uses an agent to switch between three personas to give career feedback.

Executive Coach

What's the goal?
What's the current reality?
What's the gap between them?

Economist

What's the tradeoff?
What are you giving up?
What's the opportunity cost?

Provocateur

What assumption might be wrong?
What's the contrarian view?
What would break this plan?

How to set this up

Create a custom agent (Claude Project, ChatGPT Custom GPT, or Gemini Gem) with instructions to adopt these three distinct personas. Upload the Three Lenses framework documentation so the agent understands each perspective deeply.

Level 04

Skills

Same Process and Output Format Each Time (Advanced)

Case: Combining Multiple Frameworks like Mentor Mindset + Getting Things Done using institutional logos and colors

+

The Approach: Create automated workflows (skills) that combine multiple frameworks and consistently produce formatted outputs with your institution's branding. This is the most advanced level—building repeatable processes that maintain quality and consistency.

What are Skills? Skills are pre-programmed AI workflows that execute the same process every time. Think of them as custom templates with built-in intelligence—you provide inputs, and the skill automatically applies your frameworks, follows your formatting rules, and produces consistent outputs. This is advanced because you're essentially teaching AI a complete workflow that it can repeat indefinitely with the same quality standards.

Example: Skills could be used combine all of the following

Multi-Framework Integration: Combines 2+ business/consulting frameworks (e.g., Mentor Mindset + GTD methodology)
Consistent process: Every input follows the same process and template
Institutional Branding: Automatically applies your department's logos, colors, and formatting standards

to produce consistent looking output each time.

How to build this

Required components: Upload all framework documents, create detailed instructions for the output format, include your institutional style guide and logo files. Skills requires building a Skills.md file, often also written by AI.

Time investment: Takes 2-3 hours to set up initially, but then saves hours every time you use it. Best for tasks you do repeatedly (quarterly evaluations, remediation plans, fellowship letters).

Ready for videos, newsletters, and more resources?

Keep Learning →