← Back to Sessions
U
UFT Learning LoungeProfessional Learning
Session 1 · Anatomy of a Prompt
For Facilitators Only

Facilitator Guide & Run-of-Show

A complete 60-minute plan for leading Session 1: timing, talking-point scripts, discussion prompts, and troubleshooting. By the end, participants can write a 4-part prompt and know which AI tool to start with.

📋 At a Glance

Everything you need before you walk in.

60
Minutes Total
4
Framework Parts
3
AI Tools Compared
1
Hands-on Practice
ℹ️
Session goal: Participants leave able to (1) explain in one sentence what an LLM is, (2) write a prompt using Role + Context + Task + Format, and (3) name which of ChatGPT, Claude, or Gemini they'd reach for first. Keep it practical — every example should connect to real UFT work.
⚠️
Two resource pages to share: the Participant Guide (4-part framework reference card) and the Tool Comparison page. Have both links ready to paste into chat, or print the Participant Guide in advance.

✅ Before You Start — Setup Checklist

  • Slide deck is open on screen (start at the title slide).
  • Participant Guide link is ready to paste into chat — or printed copies are on the tables.
  • Tool Comparison page link is ready for the comparison segment.
  • You have at least one AI tool (ChatGPT, Claude, or Gemini) open and logged in for the live demo — test it works first.
  • A backup demo prompt is saved in case live generation is slow or fails.
  • Timer or phone ready for the hands-on practice block.
  • You can confirm participants have access to at least one AI tool for the practice (or are paired with someone who does).

⏱️ Run of Show — Minute by Minute

60 minutes, seven phases. Adjust the practice block if you're running long.

Time Phase Facilitator Action Watch For
0:00–5:00 Welcome & Intro Title slide. Welcome the group. Ask: "Who has used ChatGPT or another AI?" by show of hands. State the goal of the hour in one sentence. Set a no-judgment tone — many will be brand new.
5:00–13:00 LLM Explainer Explain what an LLM is in plain language (see script). Use the "very well-read autocomplete" analogy. Cover what it's great at vs. where it makes things up. Don't go technical. Keep it to the one analogy.
13:00–25:00 4-Part Framework Walk through Role + Context + Task + Format one at a time (see reference below). Share the Participant Guide. Build one example prompt live with the group, adding a part at a time. Pause after each part. Ask the room to suggest the next piece.
25:00–33:00 Live Demo Paste the weak prompt into the AI, show the generic result. Then paste the full 4-part version. Compare side by side, out loud. Narrate the difference. "Notice it now sounds like us."
33:00–46:00 Hands-on Practice Participants pick one real task and write a 4-part prompt using the skeleton on the Guide, then run it. Circulate and help. Give a 2-minute warning. People who skip Format — the most common miss.
46:00–54:00 Tool Comparison Share the Tool Comparison page. Walk the table. Explain "start with what your office has." Invite anyone to try the same prompt in a second tool. Keep claims fair — no tool is "the best."
54:00–60:00 Wrap-up Recap the four parts. Run one or two discussion questions. Point to the next session. Remind them to keep the Participant Guide handy. End on a confidence note — the skill, not the tool, is the win.

🧩 4-Part Framework — Walkthrough Reference

What to teach for each part, and a line you can say.

R

Role — who the AI should be

Giving the AI an expert identity sets its vocabulary and depth.

Say: "Tell it who to be. 'You are a veteran UFT chapter leader' gives you a very different answer than no role at all."

C

Context — what it needs to know

The audience, the purpose, and any facts, tone, or rules that matter.

Say: "Who's this for? Why do you need it? What must it know? The more real detail, the more useful the answer."

T

Task — what exactly to do

One clear action with a strong verb: draft, summarize, compare, rewrite.

Say: "Be specific about the one thing you want. If you need two things, do two prompts."

F

Format — what it should look like

Length, structure, and style: email, bullets, table, FAQ, numbered steps.

Say: "This is the step people forget. Tell it the shape you want and you'll do almost no reformatting."

🎙️ Talking-Point Scripts

Word-for-word language for the key moments.

Opening (0:00)
Say this "Welcome. In the next hour you're going to learn one genuinely useful skill: how to talk to AI so it actually helps you. We'll demystify what these tools are, learn a simple four-part recipe for writing good prompts, watch it work live, and then you'll try it on something from your own job. No experience needed — by the end you'll have written a prompt you can use this week."
LLM explainer (5:00)
Say this "Tools like ChatGPT are 'large language models.' The simplest way to picture one: it's an extremely well-read autocomplete. It has read an enormous amount of text and learned to predict what words should come next. That's why it's brilliant at drafting, rewriting, and summarizing — and also why it can sometimes state something wrong very confidently. So it's a fantastic first-draft partner, but you're still the editor. You bring the judgment; it brings the speed."
Framework intro (13:00)
Say this "Here's the recipe. Four parts: Role, Context, Task, Format. Role — who the AI should be. Context — what it needs to know. Task — what exactly you want it to do. Format — what the answer should look like. Miss one and the AI guesses. Include all four and it stops guessing. Let's build one together right now."
Live demo (25:00)
Say this "Watch this. First I'll give it the lazy version — 'write an email about member benefits.' See how generic that is? Could be any organization. Now the same request with all four parts... look at the difference. Same AI, same minute. The only thing that changed was the prompt. That's the entire skill."
Launching practice (33:00)
Say this "Your turn. Pick ONE real thing from your job — an email, a summary, an explanation. Use the skeleton on your Participant Guide: fill in Role, Context, Task, Format. Write it, run it, and see what comes back. You have about ten minutes. I'll come around — wave me over if you get stuck."
Wrap-up (54:00)
Say this "Four parts: Role, Context, Task, Format. That's it. The difference between a frustrating AI experience and a genuinely useful one is almost always the prompt, not the tool. Keep your Participant Guide handy, start with whatever AI your office already has, and remember — you're the editor, never paste anything confidential. Nice work today."

💬 Discussion Prompts

Use one or two in the wrap-up, based on the room's energy.

1. Which of the four parts do you think you'll most often forget — and why?
2. Think of a task you do every week. Is it a good fit for AI, or not? How do you decide?
3. The AI sounded confident but got a fact wrong. What's your move?
4. Based on the comparison, which tool would you reach for first, and for what kind of task?

🛠️ Troubleshooting — Common Problems

❌ "I can't think of a task to practice with."
✅ Ask: "What's one email or explanation you had to write in the last two weeks?" If still stuck, offer: "Draft a welcome email for a new member."
❌ Participant writes a one-line prompt and gets a generic answer.
✅ Point at the skeleton: "You've got the Task — now add Role, Context, and Format. Add one and run it again." Show the improvement live.
❌ The AI made up a fact or sounded too confident.
✅ Great teachable moment. "This is why you're the editor. AI drafts; you verify. Never send a first draft unchecked."
❌ "Which tool should I use? Just tell me the best one."
✅ "Honestly, start with what your office already has. The prompting skill transfers to all three — that's what matters."
❌ Someone doesn't have access to an AI tool.
✅ Pair them with a neighbor for the practice, or have them write the prompt on the Participant Guide and run it later.
❌ The live demo is slow or the tool errors out.
✅ Switch to your saved backup output and narrate it. Don't let a loading spinner stall the room.

❓ Common Questions

Quick, honest answers you can give on the spot.

"Is it safe to put work information into AI?"
Treat anything you type as potentially stored. Never enter confidential member data — IDs, home addresses, anything private. General, non-sensitive drafting is fine.
"Will AI replace my job?"
It's a tool that speeds up drafting and busywork, not a replacement for judgment, relationships, or union expertise. Think of it as a very fast assistant you supervise.
"Do I need the paid version?"
No. The free tiers are plenty for learning the framework and everyday drafting. Paid plans mainly add higher limits and the newest models.
"Why did it give me a different answer than my neighbor for the same prompt?"
These tools are designed to vary their responses. That's normal — run it again or refine your prompt if you want something closer to a specific result.
"What if I don't like the answer?"
Just reply in the chat — "make it shorter," "warmer tone," "add a call to action." It's a conversation, not a one-shot.