Facilitator mode

Summer Kickoff — Facilitator Notes

Open this on your second screen. The slide deck is your projector; this is your run-of-show with the private reads, demo pre-flight, and watch-fors. Time stamps assume a clean 3:00pm start; hard stop 4:30pm.

Two-screen setup: Projector → slide deck running fullscreen (press F). Your laptop → this page. Press S inside the slide deck to open a popup speaker view if you prefer the deck's built-in speaker mode.

Pre-flight (10 minutes before)

Open Claude Desktop. Sign in to S2 account.
Open Cowork. Sign in to S2 account.
Download the Bayside demo assets to your desktop:
All five download on click. Move them to your desktop for dragging into Claude/Cowork.

Both -fixed files are pre-corrected. You reveal them for the before/after so nobody watches a file rebuild live.
Open the live poll control panel and the projector view. Pin both. Teachers join at ai.schoolinthesquare.net/poll on their phones at the start.
Set the session label. In that poll control panel, fill in Organization (S2) and Session (e.g. "2026-09-15 Principals") and hit Save. Everything submitted from then on is stamped with it, so this cohort stays separate from the last one — and the word clouds start empty instead of blending with the previous group.
Print the Pick One pledge card (PDF) — one per seat, plus ~10 spares — and put a pen on every seat. This is the Block 5 slide as a one-page handout: three lines, the "schedule it before you stand up" box, and room for notes. No card or pen, no pledge.
Open the AI Studio site, ready to show live (Level 4).
Open the sample teacher daily brief, teacher-daily-brief.html (Level 5).
(No pre-test needed for the Block 4 close any more — it's a told story now, not a live demo.)
Confirm the S2 Claude account offers a temporary/incognito chat (used for the Level 3 demo). If it doesn't, check whether memory is enabled in settings; if memory is off, a fresh regular chat works the same — just delete it after rehearsal.
Slide deck open on projector. Press F for fullscreen. Press T to start the timer when you begin.
Results dashboard open in a tab for the Block 1 usage/comfort projection.
Poll control open (linked above); you'll push the "one AI move" word cloud at the end of Block 5.

Block 1 · Welcome 3:00 — 3:10

Opening poll, why this session, format expectations.

Pace10 min

Watch-fors

Don't preamble. Start with the title slide, name yourself, move.
The two real-world failure examples (airline + lawyer) earn the room's trust early. Stay specific.
The live word-cloud poll is the only live audience moment in this block — one question. First project the results dashboard (arrival-survey usage + comfort; already collected, nothing to push). Then switch the projector to poll-present and in poll-control click "Make live" on "One task you'd love AI to take off your plate." Read three answers out loud as they land.
Optional spare in poll-control: "One word: how do you feel about AI right now?" No slide cues it — push it after the task cloud only if the room needs 30 more seconds of energy.
Set the "laptops closed" rule explicitly. Don't apologize for it.

Block 2 · Under the hood 3:10 — 3:20

Five fast beats on how LLMs work + the permission-to-be-skeptical moment.

Pace2 min per beat + 90 sec skeptic

Watch-fors

Resist depth. Each beat is 90–120 seconds. You will want to go deeper. Don't.
The "King − man + woman = queen" example is the canonical analogy for embeddings. Use it.
The skeptic slide is the most important slide in this block. Land it without defensiveness.
Specifically: "cognitive development of students" and "replacement of teacher judgment" are the two phrases skeptics need to hear named out loud.

Block 3 · Policy in practice 3:20 — 3:35

The strict positions, grade-band postures, privacy red lines.

Pace15 min

Watch-fors

Frame at the top: "the policy is set. This block is operational, not consultative."
Read the five strict positions slowly. The room needs to hear them.
Privacy red lines: this is the slide a teacher will remember if they remember only one.
If Evan or Monica is in the room, briefly acknowledge their work on the policy here.

Reference to pull up if needed

Staff AI Policy — summary page with a link to the full policy (Google Doc).
Student AI Policy — for the grade-band slide.

Block 4 · Five Levels in Action 3:35 — 4:05

The heart of the session. Six demos across five levels. 30 minutes. Buffer is thin.

If you fall behind in Block 4

Levels 4 and 5 can become "wow demos" instead of live demos. Show, don't drive.
Skip the Spanish version of the lice email at Level 1 — narrate it.
Level 2: the budget runs live (audit only — fast); the deck is a pre-made reveal. Each has an answer-key slide after it.
Closing block (Block 6) can compress from 15 to 10 minutes to recover.

Level 1 · Chat — Bayside lice email ~4 min

Demo checklist:

Switch projector to Claude.
Drag lice-email-original.html into chat.
Prompt 1 (use the deck's exact wording): "Tighten this email from Principal Henderson to families. Calm, factual, action-clear. Under 180 words. Lead with what happened, then exactly what parents should do tonight."
Read original aloud (one sentence is enough). Then read cleaned version.
Prompt 2: "Now a Spanish version for our families. Keep it in the Principal's voice; warm and personable in tone, not a literal phrase. Match the English structure."
Switch back to slides for Level 2.

Level 2 · AI works with you — budget (live), deck (reveal) ~8 min

Why the split: rebuilding a file takes a couple of minutes — dead air with a room watching. So you only run the fast audit live, and reveal the corrected files you prepared ahead of time.

Budget — live audit: drag field-day-budget.xlsx into Cowork. Before you run it, ask the room "what would you check first?" — take two guesses. Then run the audit-only prompt ("find every problem and tell me what it should be; don't change the file yet"). It streams in seconds. Then open field-day-budget-fixed.xlsx (already prepared) next to the original and walk the before/after. Advance to the "seven problems" slide.
Deck — reveal: don't run this one live. Open back-to-school-night-deck.pptx and back-to-school-night-deck-fixed.pptx side by side, walk the corrections, show the prompt you used. Advance to the "deck's problems" slide.
On the deck's voice fix: typos/math/placeholder are unambiguous and AI nails them; the voice drift is a judgment call and the one you'd eyeball yourself. That's the honest "strong assistant, not an oracle" beat.
Each demo has an answer-key slide right after it; project it and compare against what Claude caught.
Anchor sentence: "I gave it a file. It opened the file. It found the problems. I pressed every button."
Corrected budget lands at ~$5,070 total and ~$13/student (Sharpies still flagged TBD).
Watch for the buns. Six errors were planted; in the first delivery Claude found a seventh — 360 burger patties but only 320 buns. That one isn't arithmetic, it's knowing a burger needs a bun. Land it: "That's not arithmetic. That's understanding what the numbers mean." (Hot dogs match at 400/400, which is why the gap reads as real.)

Level 3 · AI knows you — Ms. Patel CLAUDE.md ~5 min

Run this in a temporary (incognito) chat. Two reasons: (1) if Claude's memory picked up Ms. Patel from a rehearsal, the "before" won't be bland and the contrast dies; (2) a clean window that has never heard of you or S2 proves the entire difference came from the file — which answers the skeptic who says "it only knows that because you've chatted with it for months." Nothing from the demo is retained afterward.

Open a temporary chat. Say it out loud: "This AI has never heard of me or this school. Watch what one file does."
First: without memory file. Prompt: "Draft a one-paragraph parent email about Jessica needing more time on the persuasive essay."
Read the generic output.
Then, same chat: attach ms-patel-CLAUDE.md. Run the same prompt.
Read the new output. Listen for the room to laugh / nod at the Otter Question of the Week.
Before moving on, say: "This email names a student and her academic work. In real life this pattern lives only in the S2 Claude account, never a free or personal tool. Jessica is fictional, so today is fine." This is the exact move teachers will try tomorrow; point them at the right place to try it.
Don't explain it. The side-by-side IS the explanation.

Level 4 · Build — the AI Studio site ~4 min

Demo checklist:

Switch to the AI Studio site. Click through a page or two they've already used today.
The punchline: "the poll on your phones an hour ago, I built that, I didn't buy it."
Honest framing: "This is the ceiling, not the floor. Most of you will not build sites. You should see that the door is open."
Do NOT promise teachers they'll build sites next week.

Level 5 · Agents — sample teacher daily brief ~4 min

Demo checklist:

Open the sample brief: teacher-daily-brief.html. It's a Bayside mock, no real data, same Ms. Patel as Level 3.
Walk one or two items the agent surfaced overnight: a student to watch, the drafted parent note, the flagged discussion thread.
Anchor: she's on the loop, not in it. It ran without being asked; she reviews, approves, or ignores.
Optional "it's real" punch: mention you run one for your own work; you don't have to show it.
Honest framing: "Most agents in 2026 still need more supervision than they save in time. You should see the shape so you recognize it when it arrives."

Closing of Block 4 — How AI fails in 2026 ~2 min

No live demo here (the old Maya/Theo trick was cut — models refuse it now, so it cost three minutes to watch nothing happen). Two slides: the Madagascar story, then the five-failure-modes summary. Anchors the room back to Block 1's airline/lawyer opener.

Run it like this:

Bridge out of Level 5 first: "That's the top of the ladder. Before you go anywhere near it, you need to know how this stuff fails — because it doesn't fail the way you'd expect."
Madagascar slide, ~60 seconds. Tell it as a story: the hidden white-text instruction, then 32 of 35 finals coming back with Madagascar in them. Then the flip — students can do this to you.
Five shapes slide. Read them, land the defense: read what comes back, check what matters, don't sign your name to anything you haven't verified.
If someone asks "can I trick it?" — the Maya/Theo passage is in the Madagascar slide's speaker notes as a back-pocket demo. Current models usually refuse it, which is itself the honest answer.

Don't chase a gotcha that reliably fools it — those are dead, and a fizzle costs you the room. The honest, sturdier lesson: the easy failures are gone; the ones left are rare, subtle, and confident. Closing line for Block 4: "The rule isn't 'AI always lies.' It's: when it's wrong, it's wrong with total confidence. So verify the things that count."

Block 5 · Pick one + partner 4:05 — 4:15

The pledge

Three lines on the printed pledge card at each seat.
One task. One AI move. One partner.
Critical: the calendar move. "Open your calendar right now. Schedule the 15-minute check-in. Before you stand up."

Partner logistics

People find their own partner, with a criterion: someone whose day looks different from theirs — different team or campus, or a different comfort level with AI — and not the person they walked in with. (Adjust the wording to the room: "different team or campus" for an admin group, "different subject or grade band" for teachers.)
Say the criterion out loud and give ~30 seconds. Nudge anyone still seated to stand and pair up so no one's stranded.
No spoken report-out. After the pledge, run the "one AI move" word cloud on the next slide instead.

Share the move — word cloud (replaces the spoken report-out)

Push the "one AI move you're going to try" question from poll-control. Everyone types theirs; read three of the more concrete ones aloud so people can borrow.

Nobody's put on the spot, and it's your cut point if you're behind, project, read one, move on.

Block 6 · Closing 4:15 — 4:30

Where everything lives (+ the honest cost)

60 seconds of honest cost first. Not "look what I built" — Level 4 already covered that. This is what they can't see: a few weeks of evenings, and plenty that didn't work first time (the budget demo shipped with a wrong number; the first cut of this deck was too long by half). Land it: "one person with a full-time job could do this — that's the bar, not 'build an app.'"
Project the AI Studio dashboard (index.html). Walk it briefly: prompts, policy, workshops, showcase.
Say the password out loud: AI@S2Training
(The old "One honest example" slide was cut — it repeated the Level 4 "I built this site with AI" beat. When a teacher leader is recruited, a real second showcase belongs here.)

Land the close

Thank you. Stick around 15 min if you want to try a prompt.
End ON time. Do not drag.

After the session

Stop the timer. Note actual end time.
(No recording promised — don't create an expectation of one.)
Email the room: link to AI Studio, password, pledge submission form, prompt library.
Collect pledges. 7 days later: send a one-line check-in email ("how's the change going?").
Note what to change before the next building's session.

Facilitator notes · v1 · Last updated June 2026

← Course page