What we're doing today

• Framing: what AI is, what it isn't, and why it matters
• Safe working: two basic rules
• How to prompt well: when to use it, and how
• Demo: two examples
• Hands-on practice: try it yourself
• Q&A and free exploration: debrief and questions

Hold this question throughout

"What's the most time-consuming, repetitive, or draining task in my week?"

We'll come back to this at the end.

The goal is to leave today with one specific thing to try with AI today or tomorrow.

This is session one

No expectation of expertise. You don't need to know how AI works to use it effectively; you do need to know what you're trying to do.

Curiosity and willingness to experiment is enough.

Today's goal is one practical win: something that saves time or reduces effort on a task you actually do.

Follow-up sessions will cover more complex use cases once there's a baseline of increased confidence and experience.

AI is a conversation partner

It's not a search engine; don't ask once and accept whatever comes back.

• The first response is a starting point, not a finished product; read it critically before using it
• You iterate: ask it to revise, correct a specific part, shorten it, or try a different angle
• Most good outputs take two or three exchanges; the back-and-forth is where most of the value is

AI transforms information

One of its most useful capabilities: taking something in one format or structure, and turning it into something different:

• A PowerPoint → a student-facing announcement email
• An email chain → a numbered action list with deadlines
• A 40-page report → a two-paragraph, bespoke summary for a specific audience

The input can be anything you can paste or upload. AI handles the reshaping.

AI extends what you can do

AI doesn't decide what to do. You do.

It handles the routine so you can focus on what requires:

• Your judgement: knowing when a response isn't right for your colleagues or working context
• Your knowledge of students, partners, and institutional relationships
• Your professional accountability for the outcome

Tasks change. Jobs don't disappear, but the job looks different when the routine parts are faster.

Task, not role replacement

"The people who learn to use AI well will be the most valuable but because they can do the same work, at higher quality, more quickly."

AI literacy is becoming a baseline professional skill, much like email or spreadsheets in earlier decades. This session is about starting to build that literacy before it feels urgent.

Useful reframe: What could I do better if AI handled the routine parts?

Rule 1: Use Copilot for work tasks

Copilot sits inside our M365 environment; it's part of the same system as Outlook and Teams.

• Covered by our institutional licence; you're not agreeing to personal terms of service
• Your data does not leave our systems and isn't used to train the model
• With ChatGPT or other tools, the terms vary and your inputs may be retained or used

If you're working with anything sensitive — student information, placement data, internal documents — Copilot is the safe default.

Rule 2: Never work on the live copy

Before you ask Copilot to help with any document or spreadsheet:

1. Save a new version first — use "Save a copy" or rename the file
2. Treat it like tracked changes — the original stays intact
3. If Copilot produces something unexpected, you can always revert

This habit matters because Copilot edits in place; there's no undo if you've closed the file.

(Copilot is good, not infallible — especially with numbers or dates. Verify before you act.)

When NOT to use AI

AI is not always faster: setting realistic expectations saves frustration.

Use AI when:

• Repetitive you do often (drafting the same type of email often)
• Reshaping something you already have (document → summary)
• It would take you a long time even though it's not complex (reading a 40-page report)

Don't use AI when you'd finish it faster by just doing it

RGID: a structure for prompts that work

The Discuss step is where most of the value is. The first response is rarely the final one.

Your task

Task A — lower barrier

Pick an email you need to write today (or one you write regularly). Use RGID to draft it in Copilot. Then use Discuss to refine it until it sounds like you.

Task B — slightly more

Paste a document, email thread, or set of notes you have in front of you. Ask Copilot to transform it: a summary, an action list, a student-facing version, or bullet points for a meeting.

While you're working

• Try a follow-up question — change one thing (tone, length, format) and see what happens. This is the Discuss step in action.
• Try removing context from your prompt — write a vague version of the same request and compare the outputs. This lands the lesson better than any explanation.
• If you hit a wall, note what you tried — we'll troubleshoot together during the debrief.

The goal is one successful, useful interaction per person. Not mastery.

Debrief

"Did anyone get something useful out of that? Did anyone hit a wall — what happened?"

Both outcomes are useful. People who hit walls usually did so for one of three reasons: the prompt lacked context, they accepted the first response without following up, or they chose a task where AI wasn't the right tool. Each of those is a lesson worth sharing with the whole group.

Your most time-consuming task

"You've been thinking about your most time-consuming tasks. Does anything you've tried today connect to that? Is there a specific task worth trying with Copilot this week?"

Even if today's session didn't produce a direct match, most high-volume admin tasks have an AI-assisted version. The next step is identifying one and experimenting — not committing to a new way of working, just running a test.

FAQ

"Can I use it for [specific task]?" Almost certainly; but try it and evaluate. If the output needs significant correction every time, the overhead isn't worth it. If it gets you 80% of the way there, it probably is.

"What if it gets it wrong?" That's why you verify before using the output, and why you always keep the original file. Treat Copilot like a capable colleague, not an authority.

"Is it keeping my data?" Within M365, Copilot doesn't use your inputs to train the model, and your data stays in our tenant. That guarantee doesn't extend to personal accounts or third-party tools.

One thing to do this week

Try using Copilot for one routine task this week.

Share what happened — a line in Teams is enough:

• What worked
• What didn't
• What you had to fix

That shared learning is how the team builds capability together.