The Prompt Ladder — Facilitator Guide

📋 At a Glance

Everything you need to know before you begin

25–30

Minutes Total

4

Levels / Rounds

Solo

Individual Work

125

Max Points

ℹ️

Core mechanic: Each participant picks ONE real task from their own job and rewrites the same prompt 4 times, adding a new ingredient each round. By Level 4, they have built a persona-powered, format-specified, constraint-guarded expert prompt — starting from a bare one-liner.

✅ Before You Start — Setup Checklist

Padlet (or Google Doc) is open and the link is ready to paste into chat — participants need this URL before Level 1
Presentation slides are on screen (Prompt Ladder deck, start at Slide 1)
Participant cards are either printed OR the URL is shared in chat
Timer app or phone is ready (you'll time each round)
You have tested the Padlet yourself — can participants post without an account? (Anonymous mode recommended)
Participants have their UFT persona from the Session 2 resource page saved and accessible — remind them before Level 4
You have a brief example prompt in your back pocket for each level, in case participants freeze

⏱️ Run of Show — Minute by Minute

Time	Phase	Facilitator Action	Watch For
0:00–3:00	Launch & Orient	Show Slide 1 (title). Distribute or share participant card link. Read the game concept aloud. Share the Padlet link in chat.	Make sure everyone can access the Padlet before starting.
3:00–5:00	How to Play	Show Slide 2 (rules). Walk through the 4 steps and point system. Emphasize: "Build on it — never start over." Answer questions.	Clarify the scenario step — participants should pick a REAL task, not an imaginary one.
5:00–7:00	Ladder Overview	Show Slide 3 (all 4 levels). Give them 60 seconds to pick their scenario and write it on their card.	Some participants freeze here. Prompt: "What is ONE thing you had to write or explain to someone last week?"
7:00–12:00	🪜 Level 1	Show Slide 4. Start 5-min timer (visible). Circulate/monitor. At 4:00, give a "1 minute left" callout. Remind them to paste to Padlet.	Participants who write a paragraph instead of a sentence — gently redirect. "Make it shorter — one sentence is the goal."
12:00–17:00	🪜 Level 2	Show Slide 5. Remind: "Don't delete your Level 1 prompt — add to it." Start 5-min timer. At 3:00, prompt: "Have you added WHO this is for?"	Participants adding too little. Encourage at least 2 context tags from the card.
17:00–22:00	🪜 Level 3	Show Slide 6. Remind them to add their own role AND a format. Start 5-min timer. Call out format options at the 2-min mark.	Participants who skip the format. This is the most common miss — nudge: "Tell it HOW to structure the response."
22:00–29:00	⭐ Level 4	Show Slide 7. Remind participants to pull up their persona. Start 7-min timer. At 3 min, remind bonus: "If you have time, try it in the other AI too." Announce voting at 1 min remaining.	Participants who haven't saved their persona — they can write a short version on the spot (2–3 sentences is fine).
29:00–35:00	Vote & Debrief	Show Slide 8. Pull up the Padlet. Read 2–3 Level 4 entries aloud. Facilitate the vote (emoji reaction or hand raise). Announce winner. Run debrief questions.	Keep energy up — celebrate the winner AND the improvement shown from Level 1 to Level 4.

🎙️ Facilitator Scripts

Word-for-word language you can use at each key moment

Opening (0:00)

Say this: "We're going to do a quick activity called The Prompt Ladder. The goal is simple: you're going to write the same prompt four times — starting bare bones and upgrading it each round. By the end, you'll have a real, work-ready prompt that you can actually use at the UFT. No pairs, no groups — just you and your AI tool. Pick a task you genuinely do at work, and let's climb."

Starting each level

Level 1: "Level 1 — The Bare Ask. I want you to write the simplest version of your task you possibly can. One sentence. No context, no formatting, no role. Just ask for the thing. You have 5 minutes — go."

Level 2: "Level 2 — Task plus Context. Don't delete what you wrote. Add to it. Tell the AI who this is for, why it's needed, and what needs to be included. At least two of those. Five minutes — go."

Level 3: "Level 3 — adding your role and a format instruction. Tell the AI who YOU are — 'I am a UFT blank-blank.' And tell it how to structure the output. Email? Bullet list? FAQ? Pick one and add it. Five minutes."

Level 4: "Last level — this is the expert round. Two moves: add your UFT persona from the resource page, and add at least one constraint — a rule the AI must follow. You have 7 minutes. And if you finish early, try it in the other AI and see what happens. Bonus points if you do."

Announcing the vote

Say this: "Okay — pencils down. Let's look at the Padlet together. I'm going to read a few of the Level 4 entries out loud. After I read them, I want you to vote — emoji react, raise your hand, or drop a name in chat — for the output that you think is most ready to actually send or use. Most useful, most on-brand, least amount of editing needed."

Wrapping up

Say this: "Here's what just happened. You started with a sentence that could have come from anyone on the internet. You ended with something that sounds like YOUR department, for YOUR audience, in YOUR voice. The difference isn't the AI — the difference is the prompt. That's the whole point of this series."

🪜 Level-by-Level Reference

What participants add at each step and what to look for

1

THE BARE ASK — 10 pts

Element added: Nothing — this is the starting baseline. Raw, context-free ask.
Expected output: Generic. Could be from any organization. No union voice, no audience specificity.
Purpose: Creates a baseline so participants can visually see how much the prompt improves.

10 pts

2

TASK + CONTEXT — 25 pts

Element added: WHO it's for, WHY it's needed, WHAT must be included, WHERE it will be used.
Expected output: Noticeably more specific. Audience-relevant. Still may lack UFT tone or structure.
Watch for: Participants adding only one context tag. Encourage at least two.

25 pts

3

CONTEXT + TASK + FORMAT — 40 pts

Element added: Their own professional role + explicit format instruction (email, list, FAQ, table, etc.)
Expected output: Structured, scannable, immediately usable format. Starting to sound more professional.
Watch for: Participants skipping the format instruction — this is the most commonly missed step.

40 pts

4

THE FULL PACKAGE — 50 pts + VOTE

Element added: Full UFT persona (from Session 2 resource page) + at least one constraint rule.
Expected output: On-brand, audience-specific, structured, constrained — genuinely work-ready output.
Watch for: Participants without their persona saved. A short 2–3 sentence version is fine.

50+

💬 Debrief Discussion Questions

5–8 minutes. Pick 2–3 based on energy in the room.

1. Look at the Padlet — compare a Level 1 and Level 4 entry from the same person. What specifically changed?

Follow-up: Was the change what you expected, or were you surprised?

2. At which level did the AI response start to feel like it could be from the UFT — not just any organization?

Follow-up: What was the specific ingredient that made that shift happen?

3. For those who tried both ChatGPT and Claude with the same Level 4 prompt: what was different?

Follow-up: Did one feel more "UFT" than the other? Why might that be?

4. Which level produced output you could use at work this week with the least editing?

Follow-up: What would you still need to change before sending it?

5. What do you notice about the constraint you added in Level 4? Did the AI follow it?

Follow-up: What happens if the AI ignores your constraint? (Answer: add it again more explicitly, or ask it to try again.)

6. Thinking about your daily work — when would you realistically use a Level 4 prompt vs. a Level 2 prompt?

Follow-up: Not every task needs a full Level 4. When is "good enough" actually good enough?

🛠️ Troubleshooting — Common Problems

❌ "I can't think of a scenario."

✅ Prompt them: "What is ONE email, document, or explanation you had to write in the last two weeks? Start there." If still stuck: use "write an email to a new member explaining their benefits."

❌ Participant keeps starting over instead of building on the prompt.

✅ Physically show them on screen: "Take your Level 2 prompt — paste it here, then add one more sentence at the top or bottom. That's it."

❌ "My AI keeps giving the same kind of answer no matter what I add."

✅ Often means the context is too vague. Help them be more specific — "teachers" is better than "people," "open enrollment email" is better than "communication."

❌ Participant doesn't have their persona from Session 2.

✅ Direct them to the Session 2 Resource Page link in chat. If they don't have it, have them write a quick 2–3 sentence version: "You are a UFT [role] with expertise in [topics]. Your audience is [who]. Use a [tone] tone."

❌ Some participants finish all 4 levels early.

✅ Great — encourage them to: (a) try the same Level 4 prompt in the other AI, (b) add a second constraint and see what changes, or (c) help a neighbor who is stuck.

❌ Padlet is loading slowly or someone can't post.

✅ Have a backup: a shared Google Doc with pre-labeled sections (Level 1, 2, 3, 4) where participants can paste into comments.

🏆 Scoring Summary

Action	Points	Verified By
Complete Level 1 + paste to Padlet	10 pts	Padlet post visible
Complete Level 2 + paste to Padlet	25 pts	Padlet post visible
Complete Level 3 + paste to Padlet	40 pts	Padlet post visible
Complete Level 4 + paste to Padlet	50 pts	Padlet post visible
BONUS: Voted best Level 4 output by group	+5 pts	Group vote during debrief
BONUS: Finished all 4 levels before time called	+5 pts	Facilitator callout
BONUS: Tried same prompt in both ChatGPT AND Claude	+5 pts	Self-reported in chat/Padlet
Maximum possible score	125 pts

💡

Points are motivational, not evaluative. The real goal is the progression — participants seeing with their own eyes how a bare prompt becomes a work-ready asset through intentional layering.

🪜 The Prompt Ladder — Facilitator Guide