Help your team put AI into practice
In sixty minutes, your team will see three live demonstrations, practice on their own work, learn the guardrails, and choose one task to use AI on that week.
One demonstration fails on purpose, because confidence comes from knowing both what the tool can do and how it can go wrong. The full running order and two printable files are included.
For you
Session facilitator guide
Timed running order, the three demos with exact prompts, the objections table, and what to send before and after. Room to mark up afterwards.
Download Word · 14 KB
For them
Participant handout
Two pages to keep. What these tools are, the four things to try this week, the before-you-paste table, and a line to write their commitment on.
Download Word · 12 KB
Do the demos live, on real work, and let the third one fail.
A session where AI is described produces nodding and no change. A session where people watch a tool invent a citation, in the room, in front of them, produces the one habit that actually protects them.
So do not over-rehearse. Use your own half-finished documents and your own sent folder. The demos are more convincing when they are slightly rough, because rough is what the room will experience on Monday.
Four things to do first
- Send one email a week out. Ask everyone to bring a real task they are avoiding, plus five emails they have actually sent. Say it twice, because half of them will not do it.
- Answer the tool question yourself. Which tools your organization has approved, whether any is cleared for confidential data, and who to ask when unsure. If you cannot answer those three, the session will stall on them in the first ten minutes.
- Print the handout and the two checklists. One per person. On paper, not on screen. The checklists are here.
- Set up your own screen. Tool open, logged in, with a real document ready to work on.
Running order
| Time | Segment | What you are doing |
|---|---|---|
| 0:00 | What these things actually are | Five minutes, no slides. Very good at shape, unreliable with facts. |
| 0:05 | The three weeks | Name the pattern so nobody quits in week two. |
| 0:10 | Demo 1: make it sound like you | Live. Use your own sent emails. |
| 0:20 | They do it | Hands on keyboards. Their emails, not yours. |
| 0:28 | Demo 2: criticize my finished work | Live. Use a real document of yours. |
| 0:35 | Demo 3: watch it fail | Live. Ask for citations. Look them up in the room. |
| 0:43 | Guardrails | Walk the three lists. Hand out the checklist. |
| 0:52 | One commitment each | Round the room. One task, this week. |
| 0:58 | Where to go next | Point at the site. Close. |
That is 58 minutes of a 60-minute slot, deliberately. If you are behind at 0:35, cut the hands-on block rather than demo 3. Demo 3 is the point of the session.
What to put on screen
01 Make it sound like you
Paste five of your own sent emails. Not theirs, not made-up ones.
Here are five emails I have written. Describe my writing voice back to me: how I open and sign off, sentence length, how direct I am, how formal I am, and what I never do. Then turn that into a short set of instructions you could follow to write in my voice. [paste five emails]
What to say while it runs: this is the whole trick. Generic AI writing is recognizable, and faculty will think less of you for sending it. Your sent folder is the fix. Then use the profile to draft one real email you owe somebody, and show them the two-version trick: ask for one direct and one softer, then pick.
02 Criticize my finished work
Use a document you actually wrote and were happy with. It has to be yours, or the demo has no weight.
Name the three weakest things about this document. Don't tell me what works. Assume a doubtful reviewer who thinks it overstates its case. What do they go after first? [paste your document]
The teaching point is not that it is right. Ask the room: out of three criticisms, how many are useful? One or two is a realistic hit rate, and that is a second reader available at 10pm. Then show the other half: ask "is this good?" and let them watch it tell you the document is good.
03 Watch it fail
This is the one that changes behavior. Pick a topic you know thoroughly.
Give me three peer-reviewed references on [a topic you know thoroughly], with authors, journal and year.
Now look them up, live, in front of everyone. Depending on the tool and whether it searched the web, the references may be real, may not support the claim, or may not exist. The point is to verify, not to guarantee a failure on demand.
If all three are real and relevant, say so. The checking routine still matters because a successful attempt does not validate future references.
The objections you will get
Short, honest answers work better than reassurance. This room can tell when it is being sold something.
- "Is this allowed?"
- Answer it factually, with what you found out before the session. If there is no policy yet, say so, and say that finding out is on the list. Do not guess at this one.
- "Isn’t this cheating?"
- The accountability does not move. You are still the author and you still check the content. Drafting with a tool is closer to using a template than to outsourcing.
- "It got it wrong when I tried it."
- Often a missing-context problem. Ask what source material, audience, constraints and examples they supplied before judging the workflow.
- "I don’t have time to learn this."
- Agree, then narrow it. The faculty email recipe pays back inside one afternoon. Nobody has to learn anything else today.
- "Will this replace my job?"
- Take it seriously rather than brushing past it. It drafts and reorganizes. It cannot run a needs assessment, judge whether content is valid, or hold a hard conversation. Then be honest that nobody can promise how this looks in five years.
- "What about patient data?"
- This is the right question. Go to the guardrails checklist and walk it. Reward whoever asked.
One commitment each, out loud
Go round the room. Each person names one real task they will use this on within the week. Not a category. A specific task with a deadline attached.
Write them down and send the list round afterwards. This is the single thing that separates a session people enjoyed from a session that changed anything, and it takes six minutes.
Then put a 15-minute follow-up in the calendar for two weeks out. Not another session. Fifteen minutes, one question: did you do the thing?