Run an AI lunch-and-learn for colleagues

The problem

You've been experimenting with AI and found it useful, but your colleagues haven't started yet. Some are curious but don't know where to begin. Others are sceptical or anxious about AI. A few think it's just hype. You want to share what you've learned in a way that's practical, not preachy, and gets people actually trying things rather than just listening to a presentation.

The solution

Run a hands-on lunch-and-learn session where colleagues try AI themselves rather than watch slides about it. The secret is immediate, relevant tasks: bring your charity's actual documents (anonymised) and have people work on real problems. Sixty minutes of doing beats three hours of theory. Keep it informal, acknowledge limitations honestly, and focus on 'this could save you time' rather than 'this will change everything'.

What you get

A 45-60 minute session where colleagues leave having actually used AI on a work-relevant task. They'll have a Claude or ChatGPT account set up, understand basic prompting, know what AI is good and bad at, and have one technique they can use tomorrow. You'll have identified the curious early adopters and addressed common concerns.

Before you start

• Your own experience using AI (you need to demonstrate, not just describe)
• Sample documents or data from your charity (anonymised)
• A room with wifi and colleagues who can bring laptops or phones
• Manager support for running the session during work time
• NOTE: Brief participants that free AI tiers (ChatGPT Free, Claude Free) may use conversations for model training. Advise them to assume anything typed is no longer private, and never enter real beneficiary names, case details, or confidential data during practice.

When to use this

• Colleagues are curious about AI but haven't started
• You want to build grassroots AI capability before a formal rollout
• Leadership has approved AI use but adoption is slow
• You've found useful AI techniques and want to share them

When not to use this

• You haven't used AI enough yourself to demonstrate confidently
• Your organisation has banned AI use
• Colleagues have no access to devices during work

Steps

1. 1

   Plan your demo tasks

   Choose 2-3 tasks that are immediately relevant to attendees. Good options: summarising a long document everyone knows, finding themes in feedback data, drafting a standard email. The task should take 5 minutes and produce a useful output. Avoid tasks that require explanation.

2. 2

   Prepare anonymised materials

   Gather real materials from your charity work: a board paper everyone has seen, anonymised survey responses, a long policy document. Real materials are more compelling than made-up examples. Remove any personal data.

3. 3

   Structure your session (45-60 mins)

   Suggested structure: 5 mins - what AI is (briefly), 10 mins - live demo of a real task, 20 mins - everyone tries it themselves, 10 mins - discuss what worked/didn't, 5 mins - what to try next and any policies. The middle section where people try things is the most important.

4. 4

   Handle account setup in advance

   The biggest friction is account creation. Send instructions beforehand asking people to create a Claude or ChatGPT account before the session. Have a backup plan for those who didn't (they can watch and try later, or pair with someone who has an account).

5. 5

   Demonstrate honestly

   Show a real task, including when AI gets things wrong. "Watch, this is useful" - demonstrate a good result. "But look, here it made something up" - show a hallucination. Honesty builds trust. Demonstrate the review process you use.

6. 6

   Give people a specific task

   Don't say "now try it yourself" - give a specific task with specific text to paste. "Copy this survey feedback, paste it into Claude, and ask it to find the top three themes." Specific beats open-ended for beginners.

7. 7

   Address concerns directly

   Common concerns: "Is this allowed?" (reference your policy), "Will it replace me?" (no, it's a tool), "What about data privacy?" (explain what not to input), "Isn't it just hype?" (show real time savings). Don't dismiss concerns; address them.

8. 8

   End with next steps

   Give people one thing to try this week: "Try summarising your next long email before responding" or "Before your next meeting, ask Claude to suggest questions about the agenda." Follow up a week later to see how it went.

Example code

# AI Lunch and Learn - Running Order

## Before the session
- [ ] Send calendar invite with "please create a free account at claude.ai OR chatgpt.com before the session"
- [ ] Prepare anonymised documents (survey feedback, long report, etc.)
- [ ] Test wifi and screen sharing
- [ ] Print quick reference cards if possible

---

## 0:00-0:05 - Welcome and ground rules (5 mins)
"This is hands-on - you'll try things yourself, not just watch me.
There are no stupid questions.
What happens in this room stays in this room (your early prompts might be clumsy!).
We'll cover what AI can do, try it ourselves, and discuss what this means for our work."

## 0:05-0:15 - Quick demo: real task (10 mins)
Live demonstration (not slides):
1. "Here's that 15-page board report from last month"
2. Paste into Claude: "Summarise this board report in 5 bullet points, highlighting any decisions made and actions required"
3. Show the result: "That took 10 seconds. Would have taken me 20 minutes."
4. Point out any errors: "Notice it said X - that's not quite right. This is why we always review."

## 0:15-0:35 - Everyone tries it (20 mins)
Hand out the task sheet (see below).
Walk around helping people who are stuck.
Encourage people to try variations.
"If you get an error, that's normal - try rephrasing."

## 0:35-0:50 - Discussion (15 mins)
- "What surprised you?"
- "What didn't work as well as expected?"
- "Where could you see using this in your work?"
- "What concerns do you have?"

Address the common concerns:
- Data privacy: "Never paste beneficiary names or personal details"
- Job security: "This is a tool, like email was. It changes how we work, not whether we work"
- Quality: "Always review outputs. AI is a first draft, not final copy"

## 0:50-0:55 - What's allowed and next steps (5 mins)
- Reference the AI policy (or say "we're working on one")
- "Here's one thing to try this week: [specific task]"
- "Questions? Email me or grab me for a chat"

## 0:55-1:00 - Informal chat / overrun
Some people will want to keep talking. That's good.

# AI Hands-On Tasks

## Task 1: Summarise something (5 mins)

1. Go to your AI tool (claude.ai or chatgpt.com - you should already have an account)
2. Copy the text below (or use the document provided)
3. Paste it and add: "Summarise this in 5 bullet points"
4. Read the result - is it accurate?

[Paste your sample text here - a paragraph from a report, policy, or email]

---

## Task 2: Find themes in feedback (5 mins)

1. Copy these survey responses:
[Paste 5-10 anonymised survey responses here]

2. Paste them into Claude with:
"What are the main themes in this feedback? List the top 3 with a quote illustrating each."

3. Do you agree with its analysis?

---

## Task 3: Draft something (5 mins)

Choose one:

A) "Draft a short email declining a meeting politely. I'm busy that day but open to rescheduling."

B) "Write 3 social media posts about [your charity's recent news]. Keep them under 280 characters each."

C) "Suggest 5 questions I should ask when interviewing candidates for a [role] position."

---

## Tips for better results

- Be specific: "Summarise in 5 points" not just "summarise"
- Give context: "I'm a charity fundraiser" helps
- Ask for format: "As bullet points" or "in a table"
- Iterate: "Make it shorter" or "More formal please"

---

## What to try this week

Pick ONE thing to try before Friday:
- [ ] Summarise a long email before responding
- [ ] Ask for feedback on something you've written
- [ ] Generate ideas for a problem you're stuck on

Questions? Ask [facilitator name] or email [address]

Tools

Resources

Part of this pathway