Run a practical AI lunch and learn in 45 minutes with a safe demo, mixed-skill facilitation plan, follow-up kit, and course decision guide.
run an AI lunch and learn at workAI training for teamsAI workshop agendaworkplace AI training
Key Takeaways
The points worth keeping
Use a 45-minute agenda built around one useful workplace workflow, not a tour of AI features.
Demonstrate with fictional or public information; never paste confidential, personal, regulated, or customer data into an unapproved tool.
Give participants a prompt template, a small practice challenge, and a way to evaluate quality after the session.
Facilitate for mixed skill levels by separating the common outcome from optional extensions.
FAQ
Questions this topic usually raises
What should an AI lunch and learn cover?+
Cover one practical, low-risk workflow rather than a broad overview of AI. A strong session includes a clear goal, a safe fictional demonstration, guided practice, output review, and a follow-up challenge. Participants should leave with a reusable prompt pattern, an understanding of data boundaries, and a specific task they can practise under the organisation’s approved-tool and review policies.
How long should an AI lunch and learn be?+
Forty-five minutes is enough for an introduction and guided first practice. Reserve five minutes for the goal and safety rules, ten for context, ten for a live demonstration, twelve for participant practice, six for comparing outputs, and two for the next step. Longer sessions are useful when the workflow requires substantial domain context or multiple practice rounds.
How can I demonstrate AI without exposing confidential data?+
Use fictional, synthetic, or genuinely public information and label it clearly during the session. Do not paste customer records, personal information, credentials, unpublished plans, legal material, or regulated data unless the organisation has explicitly approved that use and the tool’s controls are understood. Demonstrate the review process so participants learn both what the tool can do and what it cannot verify.
If you want faster execution, open the prompt library. If you want a bigger decision, open the role guides or the course catalog.
Is there a guide-first path before buying?
Yes. Start with the guide hub, then use the sample lesson path or the prompt library before committing to membership.
How do I avoid random browsing?
Choose the next step that matches your job to be done, not the most popular page.
Put it into practice
ChatGPT for Beginners — 2026 Edition
Free course. Use ChatGPT the right way from day one: real conversations with follow-ups, the 3-part prompt, uploading files and photos, checking facts on the web, and five everyday wins. Real screen demos, five short lessons.
5 lessons. The complete course is free with your account. No card required.
Lifetime access is one payment with no renewal. Membership is US$10/month or US$80/year. AI tools and their credits may cost extra.
Choose team training for shared adoption and self-serve courses for individual depth, flexibility, or role-specific learning.
Should a company choose team training or self-serve AI courses?+
Choose team training when the priority is shared behaviour, safe-use rules, cross-functional workflows, or live discussion. Choose self-serve courses when learners need flexible schedules, role-specific depth, or repeated practice. Many organisations benefit from both: a short facilitated session to align the team, followed by courses and a practice challenge that build durable individual capability.
How do I measure whether the session worked?+
Measure more than attendance or satisfaction. Ask participants to complete a small practice task, identify an error or limitation in an output, and name a realistic next use case. Follow up after 30 days on practice completion, time saved, quality or rework, unresolved risks, and whether participants need deeper training. Avoid claiming productivity gains without a defined baseline and review method.
More AI Learning articles
Keep reading on the same topic with adjacent guides.
A good AI lunch and learn is not a compressed conference talk. It is a guided first practice. Participants should leave knowing what to try, what not to share, and how to judge whether the output is useful.
Use this agenda:
Time
Activity
Outcome
0–5 minutes
Set the goal and safety rules
Everyone knows the workflow and data boundaries
5–10 minutes
Show the before state
Participants see the cost of the current process
10–20 minutes
Run the live demonstration
The group sees prompting, checking, and revision
20–32 minutes
Guided individual or pair practice
Each person adapts the workflow to a safe example
32–38 minutes
Compare outputs and tradeoffs
The group learns that quality requires review
38–43 minutes
Introduce the 30-day challenge
Participants have a next step
43–45 minutes
Survey and call to action
You collect feedback and point to deeper learning
Before inviting people, define one sentence that describes the session: “By the end, you will be able to turn rough meeting notes into a clear action plan while protecting sensitive information.” That constraint prevents the event from becoming a general discussion about AI.
Pick a workflow people can use tomorrow
Choose a process with four properties:
It happens often enough to matter.
The first draft is easy for a human to review.
A fictional or public example can demonstrate it safely.
The output has a clear quality standard.
Good lunch-and-learn topics include summarising a public report, converting notes into an action tracker, drafting a project update, creating interview questions, or turning a policy into a short FAQ. Avoid workflows that depend on confidential customer records, legal conclusions, medical advice, employment decisions, or unsupervised external communication.
A useful workflow has this shape:
Input: notes, a brief, a public document, or fictional records.
Transformation: summarise, classify, compare, draft, or structure.
Review: verify facts, assumptions, missing context, and tone.
Action: edit, approve, store, or share through the normal process.
This framing also makes the session course-oriented. Participants are learning a transferable pattern rather than memorising a single prompt.
A no-confidential-data demo script
Use a fictional scenario such as a team planning a product webinar. Prepare a short block of invented notes before the session:
Webinar planning notes: draft date is 14 October. Marketing owns the landing page. Product will provide a 10-minute demo. Open questions: speaker confirmation, accessibility review, registration target, and follow-up email. Risk: the demo environment may not be ready. Next planning meeting: Friday.
Tell participants explicitly that these notes are fictional. Then demonstrate the workflow in three passes.
Pass 1: Ask for structure
You are an operations assistant. Turn the fictional webinar notes below into a table with these columns:
- action
- owner
- due date or timing
- dependency
- open question
Do not invent owners or dates. If information is missing, write “not specified.”
Notes:
[PASTE THE FICTIONAL NOTES]
Pause and ask the group what the output did well and what it could not know. This teaches an important habit: a polished answer can still contain gaps.
Pass 2: Add a quality standard
Review the action table against these rules:
1. Separate confirmed information from assumptions.
2. Flag missing owners and dates.
3. Identify one risk that needs a decision.
4. Do not add facts that are absent from the notes.
Return:
- issues found
- questions for the project lead
- a revised table
Pass 3: Make it usable
Draft a concise project update for the planning team using the revised table.
Requirements:
- maximum 120 words
- use plain language
- include progress, open questions, risks, and next meeting
- label uncertain information as “to confirm”
- do not claim that any task is complete unless the notes say so
Explain that the human remains accountable for checking the notes, owners, dates, and tone. The tool can accelerate drafting; it does not validate reality.
Facilitation checklist for mixed skill levels
Use this checklist before and during the session:
Send the goal, agenda, and data rule at least one day ahead.
Confirm which AI tools are approved by the organisation.
Prepare a fictional example and a completed reference output.
Test the prompts in the exact tool and account type participants will use.
State that no confidential, personal, customer, financial, health, legal, or unpublished company data should be pasted into the demo.
Offer a copy-and-edit prompt for beginners.
Offer an extension task for experienced users, such as adding evaluation criteria.
Pair people by workflow familiarity, not only by technical confidence.
Give quiet participants a written practice option instead of requiring live sharing.
Ask participants to identify one failure or uncertainty in the output.
Reserve time for the next action and feedback survey.
For beginners, explain the prompt in plain language and let them change only one variable at a time. For experienced users, ask them to improve the specification, test a counterexample, or compare two outputs against a rubric. Both groups can work on the same workflow without forcing everyone into the same level of detail.
Worked example: from notes to a reviewed action plan
Suppose the model produces this draft:
Action
Owner
Timing
Dependency
Open question
Create landing page
Marketing
Not specified
Speaker details
Who approves copy?
Provide product demo
Product
Before 14 October
Demo environment
Is the environment ready?
Complete accessibility review
Not specified
Before launch
Draft page
Who is responsible?
Set registration target
Not specified
Not specified
Planning decision
What target should be used?
The useful lesson is not that the table looks neat. It is that the missing information is visible. A participant can now send three precise questions instead of forwarding a vague summary:
Who owns the accessibility review?
Who approves the landing-page copy?
What registration target should the team use, and by when should it be agreed?
Ask the group to score the draft against four criteria: factual accuracy, completeness, clarity, and actionability. A simple 1–5 scale is enough. If nobody can explain why an answer received a high score, the rubric needs improvement.
The follow-up kit: turn one session into practice
A lunch and learn has limited value if participants never repeat the workflow. Send a small kit immediately afterward.
1. Prompt pack
Include three reusable templates:
Structure: “Turn this material into [format]. Preserve known facts. Mark missing information as [label].”
Critique: “Review the draft against [criteria]. List unsupported claims, omissions, ambiguities, and risks.”
Adaptation: “Rewrite this for [audience] in [format] and [length]. Keep these facts unchanged: [facts].”
Tell learners to replace the bracketed fields, add context, and keep a human review step. A prompt pack should be short enough to use, not a library nobody opens.
2. A 30-day practice challenge
Set one small activity per week:
Week 1: Use the workflow on a fictional or public example and record one correction.
Week 2: Apply it to a low-risk internal task after checking the organisation’s policy.
Week 3: Compare the AI-assisted process with the old process for time, quality, and rework.
Week 4: Share one successful pattern, one failure, and one proposed guardrail.
The goal is not to maximise AI usage. It is to find repeatable work where assistance improves the result without weakening accountability.
3. A short evaluation survey
Ask five questions using a 1–5 scale:
I can explain the workflow shown today.
I know what information I should not enter into an AI tool.
I can judge whether the output needs correction.
I have a realistic task on which to practise.
I would benefit from deeper training on this topic.
Add two open questions: “What would you use this for?” and “What would make you hesitate?” Review the answers for adoption barriers, not just satisfaction.
When team training beats self-serve courses
A live session is a starting point, not always the best complete solution. Use team training when people need a shared operating model, such as approved tools, review responsibilities, common prompt patterns, or a workflow that crosses departments. Live facilitation is also valuable when leaders need to model safe behaviour or when the team has uneven confidence.
Self-serve courses are often better when learners need flexibility, role-specific depth, repetition, or a structured path they can complete asynchronously. They also suit distributed teams with different schedules and individuals who want to revisit fundamentals privately.
Situation
Better first choice
Why
The team needs a common safe-use baseline
Team session
Shared language and immediate questions
Learners have different roles and goals
Self-serve courses
Personalised pace and depth
A workflow spans several departments
Team training plus practice
Decisions and handoffs need alignment
One person wants to build foundational skill
Self-serve course
Efficient, repeatable learning path
Adoption is blocked by policy or trust concerns
Facilitated training
Concerns can be surfaced and addressed
The organisation wants durable capability
Both
Live alignment plus continued practice
A practical path is to run the 45-minute session, inspect the survey, and then direct people to relevant learning paths in the courses. Let learners compare options, including pricing, before committing to a larger programme. Use the contact page when the team needs a tailored plan.
Common mistakes and tradeoffs
Trying to cover every AI capability. Breadth feels impressive but reduces transfer. One workflow with review beats ten disconnected examples.
Using realistic confidential data. This creates avoidable privacy, security, and trust risks. Fictional data may feel less exciting, but it makes the safety rule concrete.
Showing only the successful output. Participants then assume the tool is reliable. Show a missing owner, unsupported assumption, or awkward tone and demonstrate correction.
Measuring attendance as success. Attendance indicates reach, not behaviour change. Track practice completion, quality improvements, and unresolved concerns.
Ignoring tool access. A workflow cannot spread if people lack approved accounts, permissions, or a clear way to request access. Check this before the event.
Limitations and assumptions
This plan assumes participants can access an organisation-approved AI tool and that the session is introductory rather than a replacement for security, privacy, legal, compliance, or role-specific training. The examples use fictional information and do not prove that a model will be accurate for your work. Tool capabilities, retention settings, approved-use policies, and user interfaces vary by organisation and may change. Before applying any workflow to real work, confirm the applicable policy, remove unnecessary sensitive information, verify important outputs against authoritative records, and keep an accountable human decision-maker in the loop.
For more practical ideas, browse the blog, then choose whether your team needs a focused workshop, a structured course, or both. The best next step is the smallest safe workflow you can practise this week.
