Trainer notes

Answer keys for the activities

From the speaker notes of each activity slide.

Card answer sheet

The answers to the printable cards, in the order of your card list. The same sheet can be printed from the Cards page.

AI or not AI?

Jargon matching – word cards

Jargon matching – library example cards

Token cards – Part 1: word cards

Token cards – Part 2: text cards

Token cards – Part 2: token number cards

Speaker notes for every slide

Text on each slide, then the speaker notes. Source of truth for layout is the slide HTML.

Welcome, goals and the plan for today

Slide 1: AI at Work (cover)

• BIBLIOTHEEK WESTLAND · MANAGEMENT WORKSHOP
• AI at Work
• A practical, friendly introduction to AI and Microsoft 365 Copilot
• Fahriddin Salaydinov · 10.10.2026 · Westland, Netherlands

Speaker notes: Welcome everyone as they arrive. Make sure each person has their laptop open, is signed in to Microsoft 365, and can see the Copilot icon in Outlook and Word. Set the tone from the first minute: no question is too basic, nobody here needs a technical background, and we will go at the group's pace. With only three people, this is closer to a guided conversation than a lecture – invite interruptions.

Slide 2: By the end of today, you will… (goals)

• By the end of today, you will…
• 1
• Understand
• what AI is – and what it is not – in plain words
• 2
• Talk the talk
• know the key AI words you hear in meetings and the news
• 3
• Use Copilot
• write good prompts and save time on real library tasks – safely
• AI at Work · Welcome

Speaker notes: Three simple goals. Stress that the third one is the main event: most of our time is spent with hands on keyboards. Ask: "Is there anything you are hoping to get out of today that isn't here?" Write any answers down and come back to them in the wrap-up.

Slide 3: Our plan for today (agenda)

• Our plan for today
• 1 What is AI?
• 4 Hands-on with Copilot
• 2 AI jargon, made simple
• 5 Using AI safely
• 3 How to ask: prompting
• 6 My first AI task
• We learn a little, then try it straight away. Questions are welcome at any moment.
• AI at Work · Welcome

Speaker notes: Walk through the six blocks. Mention when the break will be and where coffee and toilets are. Explain the rhythm: a short explanation, then an activity, every time. Reassure the group that blocks 1 to 3 are short and the real practice happens in block 4.

Slide 4: AI in my day (icebreaker)

• LET'S START
• AI in my day
• Each of us shares:
• 1. One task at work I wish took less time
• 2. One thing I have heard about AI – good or worrying
• AI at Work · Welcome · Activity

Speaker notes: Go first yourself to model a short, honest answer. Write each person's "task I wish took less time" on a flip chart or sticky note – you will come back to these tasks in block 6, where each person tries Copilot on their own task. Take worries seriously and don't argue with them; say you'll cover them in block 5 (safe use). Typical worries: "it will replace jobs", "it makes things up", "what happens to our data".

What is AI – a plain-language picture

Slide 5: What is AI? (s1)

• Block 1
• What is AI?
• A plain-language picture – no maths, no code

Speaker notes: Short theory block. The aim is a mental picture they can keep, not technical understanding. Keep it conversational.

Slide 6: A very well-read new assistant who has never worked at Bibliotheek Westland. (ai-simple)

• THINK OF AI AS…
• A very well-read new assistant who has never worked at Bibliotheek Westland.
• It writes fast and knows a little about everything – but it doesn't know your branches, your municipality or your history, and it sometimes sounds sure when it's wrong.
• AI at Work · Block 1 — What is AI?

Speaker notes: This is the key image for the whole day. Unpack it: Well-read – it has learned from a huge amount of text, so it can write, summarise and explain. New – it doesn't know Naaldwijk from Maasdijk, your agreements with Gemeente Westland or your budget unless you tell it. Assistant – it helps; you stay the manager and you sign off on the work. Sometimes sure when wrong – like a confident new colleague who would rather guess than say "I don't know". Ask: "How would you brief a new assistant on their first day?" That is exactly what a good prompt does – we'll come back to this in block 3.

Slide 7: How does it work? A smart guessing game (how-it-works)

• How does it work? A smart guessing game
• "Please return your books by the ______"
• due date 92%
• end of the month 6%
• banana 0.001%
• AI has read so much text that it can predict which words come next – very well. That is why it writes so fluently, and also why it can invent things that only sound right.
• AI at Work · Block 1 — What is AI? · Percentages are illustrative

Speaker notes: Everyone knows predictive text on their phone. Generative AI is the same idea on a giant scale: it predicts the most likely next word, again and again, until it has written a whole email or report. Ask the group to fill in the blank before you reveal the options. Key message: it doesn't "look up" facts like a catalogue; it produces likely-sounding text. That is why it is great at drafting and summarising, and why we must always check facts, names, dates and numbers. The percentages are made up for illustration.

Slide 8: What AI is good at – and where to be careful (good-careful)

• What AI is good at – and where to be careful
• Good at
• Drafting emails, letters and texts
• Summarising long documents
• Rewriting: simpler, friendlier, shorter
• Brainstorming ideas
• Explaining figures in plain words
• Be careful with
• Facts, names, dates and numbers
• Very recent or local information
• Personal data of patrons and staff
• Decisions about people
• Anything you send without reading
• AI at Work · Block 1 — What is AI?

Speaker notes: Ask the group for an example from their own job for each "good at" line – e.g. replying to a request from Gemeente Westland, summarising a long policy document or the Bibliotheekconvenant, explaining the branch figures to the Supervisory Board, drafting a vacancy text. For the "careful" column, connect back to the guessing game: because AI predicts likely words, a date or a number can be wrong but look perfect. Decisions about people (selecting candidates, performance reviews, absence cases) should always stay with a human. Block 5 covers safety in more depth.

Slide 9: AI or not AI? (act-ai-or-not)

• ACTIVITY
• AI or not AI?
• Together, sort the 10 cards into three piles:
• AI · Not AI · Not sure
• Then explain your choice for one card.
• The clue
• Does it learn from examples, or does it just follow a fixed rule?
• AI at Work · Block 1 · Cards are in the participant handout

Speaker notes: Print and cut out the 10 cards from the handout. With three people, sort them together on the table. Suggested answers (there's room for discussion – that's the point): AI = spam filter, streaming recommendations, Copilot writing an email, face unlock on a phone, translation app. Not AI = calculator, barcode scanner at self-service, automatic overdue reminder (a fixed rule: "send on day 21"). Not sure / discuss = library catalogue search (classic keyword search is not AI, but some newer discovery systems use AI to rank results) and sat-nav (route calculation is a classic algorithm; traffic predictions use AI). Key lesson: AI learns patterns from lots of examples; ordinary software follows rules someone wrote.

AI jargon explained with library comparisons

Slide 10: AI jargon, made simple (s2)

• Block 2
• AI jargon, made simple
• 24 words you will hear – explained with library examples

Speaker notes: Tell the group: you don't need to memorise these. The glossary in the handout has all 24 terms with library comparisons. The goal is that the words stop sounding intimidating.

Slide 11: The basics (jargon-1)

• The basics
• Artificial Intelligence (AI)
• Computers doing tasks that normally need human thinking
• Machine learning
• Computers learning patterns from lots of examples
• Algorithm
• A fixed set of step-by-step instructions a computer follows
• Generative AI
• AI that creates new text, images or sound
• Large Language Model (LLM)
• The "engine" behind chat AI, trained on huge amounts of text
• Training data
• All the text and examples the AI learned from
• AI at Work · Block 2 — Jargon · 1 of 4

Speaker notes: Library comparisons: AI – a catalogue that learns what to suggest, rather than following fixed rules. Machine learning – learning which books go together by looking at thousands of loans, instead of someone writing the rules. Algorithm – the shelving rules: fixed steps, same result every time (this is the difference we used in the 'AI or not' game). Generative AI – an author, not a catalogue: it writes something new instead of finding something that exists. LLM – a reader who has read millions of books but owns none of them, and can't tell you which page an idea came from. Training data – the collection it read; if the collection was old or one-sided, so are its answers.

Slide 12: How AI works (jargon-2)

• How AI works
• Natural language
• Normal human language – no codes or commands needed
• Chatbot / AI assistant
• A program you talk to in everyday language
• Token
• The small pieces of text (often parts of words) AI reads and writes
• Context window
• How much text the AI can keep in mind in one conversation
• Multimodal
• AI that works with text, images and sound together
• Fine-tuning
• Extra training that makes an AI better at one specific task
• AI at Work · Block 2 — Jargon · 2 of 4

Speaker notes: Library comparisons: Natural language – asking at the desk in your own words instead of in catalogue codes. Chatbot – the reference desk: you ask in normal words. Token – like syllables in a reading lesson: the AI cuts words into small pieces; these are also how AI use is measured and priced. Context window – the reading table: only so many documents can lie open at once; in a very long conversation the oldest ones slide off, so start a fresh chat for a new topic. Multimodal – a collection with books, audiobooks, films and photos in one place. Fine-tuning – a general librarian taking a specialist course in local history.

Slide 13: Tokens: how AI reads and writes (tokens)

• Tokens: how AI reads and writes
• AI doesn't read whole words. It cuts text into small pieces called tokens.
• The
• library
• in
• Na
• ald
• wijk
• is
• open
• = 8 tokens
• biblio
• theek
• con
• ven
• ant
• = 1 word, 5 tokens
• Rule of thumb: 100 tokens ≈ 75 English words. Dutch names and long compound words take more tokens.
• AI at Work · Block 2 — Tokens · Example splits: every AI system cuts words slightly differently

Speaker notes: Make it concrete: a token is a piece of a word. Common short English words like "the", "in", "is" and "library" are usually one token each, because the AI has seen them millions of times. Rare words, Dutch place names and long Dutch compound words get cut into several pieces – "Naaldwijk" into maybe three, "bibliotheekconvenant" into maybe five. The exact cuts differ per AI system; these splits are examples. Rule of thumb: 100 tokens is roughly 75 English words; Dutch text usually needs more tokens for the same message. Link back to the guessing game from block 1: AI actually predicts the next token, not the next whole word – one small piece at a time.

Slide 14: Why tokens matter to you (tokens-why)

• Why tokens matter to you
• It writes piece by piece
• AI predicts the next token, again and again. That's why it sounds fluent – and can be fluently wrong.
• Its memory is counted in tokens
• In a very long chat it can lose track of the start. New topic? Start a new chat.
• Costs are counted in tokens
• Many AI services charge per token. Useful to know when you buy or compare AI tools.
• Names can go wrong
• Words cut into pieces are easier to misspell. Always check names like 's-Gravenzande.
• AI at Work · Block 2 — Tokens

Speaker notes: Four practical consequences for managers. 1) Writing piece by piece: this is the guessing game from block 1 again, at the level of tokens. 2) Memory: the "context window" is measured in tokens. Copilot can handle long documents, but in a very long conversation the earliest parts may fade, so start a fresh chat for a new subject and ask for long documents to be summarised in parts. 3) Costs: Microsoft 365 Copilot has a fixed licence price per user, but many other AI services – and some AI agents and tools you may be offered by suppliers – charge per token or per "message". When a supplier quotes a price, ask what it is based on. 4) Spelling and counting: because AI sees pieces rather than letters, it sometimes misspells unusual names or miscounts letters and words. Always check names of people, villages and organisations.

Slide 15: Token cards (act-tokens)

• ACTIVITY
• Token cards
• Part 1 · Cut the words
• Take the 8 word cards. With a pen, mark where you think the AI cuts each word into tokens. Which word has the most pieces?
• Part 2 · How big is it?
• Put the 6 text cards in order, from smallest to biggest. Then match each one to its token number card.
• Work together. There's no exact answer for Part 1 – it's about the idea.
• AI at Work · Block 2 · Cards are in the participant handout

Speaker notes: Part 1 – likely splits (examples; they differ per AI system): "the" 1, "library" 1, "Copilot" about 2 (Cop-ilot), "subsidy" about 2 (subs-idy), "Naaldwijk" about 3 (Na-ald-wijk), "bibliotheek" about 2–3, "Honselersdijk" about 4 (H-ons-elers-dijk), "'s-Gravenzande" about 5 ('s - Gr - aven - z - ande). Lesson: common English words are cheap; Dutch names and compounds get cut into many pieces. If you have internet in the room, search for a free online "tokenizer" demo and type in a few of the words to see a real split live – use only harmless words like these, never work text. Part 2 – answers in order, rough estimates for English text (about 1.3 tokens per word; Dutch is higher): a text message ("See you at 14:00 in Naaldwijk") ≈ 10 tokens; a one-paragraph Copilot prompt ≈ 60 tokens; the MT meeting transcript from practice 2 (about 550 words) ≈ 750 tokens; a 20-page annual report (about 8,000 words) ≈ 10,000 tokens; a 300-page novel (about 90,000 words) ≈ 120,000 tokens; all the emails a manager receives in a year (about 50 a day) ≈ 2.5 million tokens. Lesson: a single document is no problem for Copilot, but "everything ever" doesn't fit in its memory at once – which is why Copilot searches for the right emails and files first, and why it helps to point it to a specific source (grounding).

Slide 16: Working with AI (jargon-3)

• Working with AI
• Copilot
• Microsoft's AI assistant inside Outlook, Word, Excel and Teams
• Prompt
• What you type or say to the AI – your question or instruction
• Prompt library
• A shared collection of prompts that work well
• Grounding
• Giving the AI your own file or email so it answers from that
• Agent
• AI that carries out several steps of a task for you
• Automation
• Software doing a repetitive task by itself, the same way every time
• AI at Work · Block 2 — Jargon · 3 of 4

Speaker notes: Library comparisons: Copilot – the reference desk built into the tools you already use every day. Prompt – the question at the reference desk; the better the reference interview, the better the answer. Prompt library – a binder of standard letters that work well; you'll get one today. Grounding – 'please answer using only this book': in Copilot you do this by pointing to a file, email or meeting. Agent – a volunteer who doesn't just answer a question but does the booking, the email and the follow-up for you. Automation – the automatic overdue reminder: same rule, every time, no person needed. Automation is not necessarily AI; it can be a simple rule.

Slide 17: Risks and responsibility (jargon-4)

• Risks and responsibility
• Hallucination
• When AI invents something false that sounds convincing
• Bias
• Unfair patterns the AI picked up from its training data
• Deepfake
• A fake but realistic photo, video or voice made with AI
• AVG / GDPR
• The privacy law that protects people's personal data
• Responsible AI
• Using AI in a fair, safe and open way, with clear agreements
• Human in the loop
• A person always checks and approves what the AI produces
• AI at Work · Block 2 — Jargon · 4 of 4

Speaker notes: Library comparisons: Hallucination – a patron confidently citing a book that doesn't exist; always check the 'citation'. Bias – a collection that only holds one point of view; the AI may repeat stereotypes it read. Deepfake – a convincing forgery in the archive; relevant for media literacy programmes and for checking suspicious messages, for example a 'voice message from the director' asking for a payment. AVG/GDPR – the rules behind your members' privacy statement; they apply to prompts too. Responsible AI – like a collection policy: agreed principles about what we do and why; this is what AI guidelines for staff are for. Human in the loop – the manager signs off: Copilot drafts, you decide. This last term is the most important rule of the day.

Slide 18: Jargon matching game (act-matching)

• ACTIVITY
• Jargon matching game
• 24
• word cards
• 24
• library example cards
• Match!
• Put each word next to its library example
• Slides off. Work together. Check with the glossary when you're done.
• AI at Work · Block 2 · Cards are in the participant handout

Speaker notes: Hide or turn off the screen so they work from memory. Shuffle the 24 word cards and the 24 library example cards from the handout. Short on time? Use only the cards from one or two of the four groups (12 pairs) – the groups are marked on the cards. Three people work as one team. When they finish, they check against the glossary handout. Celebrate getting most of them; the point is familiarity, not a test. Ask: "Which word was new to you? Which one will you use this week?"

Prompting basics with the four-part recipe

Slide 19: How to ask: prompting (s3)

• Block 3
• How to ask: prompting
• A good prompt is a good briefing

Speaker notes: Link back to block 1: "How would you brief a new assistant on their first day?" You'd say what you need, why, what to use and what the result should look like. That is all prompting is. "Prompt engineering" sounds technical, but it's really just clear instructions.

Slide 20: The prompt recipe: four ingredients (recipe)

• The prompt recipe: four ingredients
• Goal
• What do you want?
• "Write a reply…"
• Context
• Why, and who is it for?
• "…to the municipality's policy adviser…"
• Source
• Which file, email or meeting?
• "…use the figures in this thread…"
• Expectations
• Length, format, tone?
• "…formal, short, no promises."
• Only the goal is required. Each extra ingredient makes the answer better.
• AI at Work · Block 3 — Prompting · Based on Microsoft's own Copilot prompt guidance

Speaker notes: This four-part recipe – Goal, Context, Source, Expectations – is the framework Microsoft itself recommends for Copilot, so it matches the help pages they'll find later. A memory aid: "What, Why, Where from, How". Point out that the ingredients don't need labels or a fixed order; you just write them in one or two normal sentences. The cheat sheet in the handout has this recipe on one page – suggest they keep it next to their keyboard.

Slide 21: Same task, two prompts (before-after)

• Same task, two prompts
• BEFORE
• "Write a reply to the municipality."
• AFTER
• Write a reply to Sandra Koppenol – policy adviser at Gemeente Westland, who asked for our subsidy figures and about Maasdijk's opening hours. Use the figures and dates agreed in this email thread. Formal but friendly, under 200 words, and make no promises about Maasdijk.
• Goal · Context · Source · Expectations
• AI at Work · Block 3 — Prompting · Names in the examples are invented

Speaker notes: Read both prompts out loud. Ask: "Which one would you rather receive if you were a new policy officer on your team?" The "before" prompt makes Copilot guess everything, so it produces a generic email that may promise figures you don't have yet or a decision the MT hasn't made. The "after" prompt gives all four ingredients. Point out that it's still just normal English – no special codes. This is the exact task they will try themselves in practice 1 with the Outlook email thread.

Slide 22: Three habits of good prompters (habits)

• Three habits of good prompters
• Keep talking
• The first answer is a draft. Reply with "shorter", "more formal" or "add the deadline".
• Let it ask you
• Not sure what to say? End with: "Ask me any questions you need first."
• Write normally
• Full sentences, plain words. Typos are fine. No magic words needed.
• AI at Work · Block 3 — Prompting

Speaker notes: These three habits matter more than any clever trick. Keep talking: people often give up after one weak answer – show that the conversation is the tool. Let it ask you: a great way to start when you're unsure; Copilot will interview you, like a good reference librarian. Write normally: older learners sometimes worry they need special syntax or perfect spelling. They don't. Also mention they can type "please" – it's fine, and it doesn't hurt.

Slide 23: Fix this prompt (act-fix-prompt)

• ACTIVITY
• Fix this prompt
• PROMPT A
• "Summarise this."
• PROMPT B
• "Write a vacancy text."
• PROMPT C
• "How are our numbers?"
• Each person takes one prompt and rewrites it with the four ingredients. Then we try the old and new versions in Copilot Chat and compare.
• AI at Work · Block 3 · Worksheet in the participant handout

Speaker notes: With three people, each takes one prompt. Give them a few minutes with the worksheet. Then run both versions live in Copilot Chat and compare. Example improved versions: A – "Summarise the attached MT meeting transcript for the director, who missed it. List the decisions and who does what by when, in bullet points, under 200 words." B – "Write a vacancy text for a volunteer coordinator (24 hours a week) at Bibliotheek Westland. The person will coordinate about 140 volunteers across our eight branches, mainly for the digital government help points. Warm, inclusive tone, under 300 words, with a short 'what we offer' list." C – "Using the BranchFigures table in this Excel file, compare visitors per branch from January to October. Which branch is growing and which is shrinking? Explain in three plain sentences for the Supervisory Board." Note: for A and C Copilot needs the file – point out that "source" is often the missing ingredient.

Slide 24: Coffee break (break)

• Coffee break
• Back in 15 minutes – next up: hands-on with Copilot

Speaker notes: Move this slide wherever the break fits best for your group. Before the break, ask participants to check that they can open the four practice files (or have received the email thread in Outlook), so the hands-on block starts smoothly.

Hands-on practice in Microsoft 365 Copilot

Slide 25: Hands-on with Copilot (s4)

• Block 4
• Hands-on with Copilot
• Four management tasks: the municipality, the MT meeting, the figures and a project proposal

Speaker notes: This is the heart of the workshop. With three people, everyone does every practice side by side rather than rotating between stations. Walk around and help. Let people go at their own speed – if one person finishes early, they can try the "extra" prompts on the exercise sheet. Remind them: Copilot gives a different answer each time, so their results will not match each other's exactly – that's normal.

Slide 26: Where to find Copilot (where-copilot)

• Where to find Copilot
• App
• Look for…
• Good for
• Outlook
• "Summarise" at the top of an email; Copilot when replying
• Long threads, replies
• Word
• The Copilot button on the Home tab
• Drafts, summaries, rewriting
• Excel
• The Copilot button on the Home tab
• Questions about figures, charts
• Teams
• Copilot in a meeting or chat
• Meeting recaps, catching up
• Copilot Chat
• The Microsoft 365 Copilot app or Copilot in Teams
• Anything – ideas, questions
• AI at Work · Block 4 · Button names and places can change with Microsoft updates

Speaker notes: Show each place live on the big screen before they start. Microsoft regularly moves buttons and renames things, so check your own screens the day before the workshop and adjust this slide if needed. Tip for Excel: Copilot only works on files saved in OneDrive or SharePoint with AutoSave on, and works best when the data is formatted as a table – the practice file already is. In Teams, meeting recaps need transcription or Copilot turned on during the meeting, which is why we practise the meeting task with a transcript in Word instead.

Slide 27: Answer a request from the municipality (prac-outlook)

• PRACTICE 1 · OUTLOOK
• Answer a request from the municipality
• Open the email "Input for 2026 subsidy report"
• Click Summarise at the top
• Click Reply, then Copilot, and type the prompt
• Ask for one change: shorter, more formal…
• Check figures and dates. Don't send!
• TRY THIS PROMPT
• Write a reply to Sandra Koppenol at Gemeente Westland. Use what Marieke and Jeroen agreed in this thread: when we send which figures, how we explain Maasdijk, and the meeting time. Formal but friendly, under 200 words, no promises about Maasdijk.
• AI at Work · Block 4 · Practice file 1 · People and figures are invented

Speaker notes: Before the workshop, forward practice file 1 to each participant as an email so the whole thread is in their inbox. A good summary shows three requests (figures by 27 November, an explanation about Maasdijk, the meeting on 8 December) and what was agreed internally. A good reply: January–October figures plus a forecast by 27 November and final 2026 figures by 31 January; Maasdijk is being analysed and the MT decides on 1 December, and we will inform her right after; the meeting is confirmed for 8 December at 14:00 at the town hall in Naaldwijk. Check together: did it promise the full-year figures too early? Did it present the Maasdijk change as decided? Is the time right? This prepares them for "spot the mistake" in block 5. Remind them firmly not to send the email.

Slide 28: Turn the MT meeting into actions (prac-meeting)

• PRACTICE 2 · WORD
• Turn the MT meeting into actions
• Open practice file 2: the MT meeting transcript
• Click Copilot on the Home tab
• Type the prompt in the Copilot pane
• Then ask: "Write a short staff update about the decisions."
• Compare with the transcript
• TRY THIS PROMPT
• Summarise this MT meeting for the director, who missed it. Give me: the decisions made, a table of actions with who and by when, and the open questions for 1 December. Keep it to one page.
• AI at Work · Block 4 · Practice file 2 · People and figures are invented

Speaker notes: Expected decisions: start recruiting a volunteer coordinator now; the collection specialist vacancy is on hold until the budget is clear. Expected actions: Anouk – three budget scenarios for 2027 by 20 November (gap about €95,000), for the MT on 1 December and the Supervisory Board on 10 December; Lotte – vacancy text by 16 November, plus an analysis of absence (6.8% against a 5% target); Fatima – results paragraph on digital inclusion for the subsidy report by 25 November, plus a first idea for expanding to other branches; Ruud – meeting with the school board about the Honselersdijk space in January (agreement ends 31 July 2027); Marieke – one-page AI guidelines for staff by 1 December. Open for 1 December: the Maasdijk opening hours and the budget scenarios. Follow-up question for the group: is the staff update suitable to send, or does it reveal things (like the budget gap) that the MT wouldn't want to share yet? A nice judgement discussion for managers. Mention that in real Teams meetings with transcription switched on, Copilot can do this directly – same prompt.

Slide 29: Ask questions about the branch figures (prac-excel)

• PRACTICE 3 · EXCEL
• Ask questions about the branch figures
• Save practice file 3 to OneDrive
• Click Copilot on the Home tab
• Ask the three questions, one by one
• Check one answer by looking at the table yourself
• TRY THESE PROMPTS
• 1. Which of our eight branches had the most visitors, and which the fewest?
• 2. Show visitors per month for each branch as a line chart.
• 3. Which branch is growing and which is shrinking? Explain in three plain sentences for the Supervisory Board.
• AI at Work · Block 4 · Practice file 3 · Figures are invented

Speaker notes: Excel is where people feel least confident, so go slowly. If the Copilot button is greyed out, the file is probably not saved in OneDrive or AutoSave is off. Expected findings: Naaldwijk has the most visitors and Honselersdijk the fewest; Wateringen is growing; Maasdijk is clearly shrinking, which links to the municipality email in practice 1; summer months are quieter everywhere. Extra prompts for fast finishers: on the "Budget 2026" sheet – "Which budget lines have spent more than 80% so far?" (Answer: personnel, collection, housing and facilities, and ICT and systems; ICT is highest at about 92%.) On the "Staff" sheet – "Which teams have vacancies, and where is sickness absence above 5%?" Stress checking: ask them to verify one number by looking at the table themselves. Note: the IDO total (1,146) matches the figure in the email thread and the project notes – a nice way to show that Copilot's answers can be checked against other sources.

Slide 30: From rough notes to a project proposal (prac-event)

• PRACTICE 4 · WORD OR COPILOT CHAT
• From rough notes to a project proposal
• Open practice file 4: notes for "Digitaal Westland"
• Open Copilot and type the prompt
• Then ask for a short press release
• Then: a LinkedIn post for our partners
• TRY THIS PROMPT
• Using these notes, write a one-page project proposal for Gemeente Westland. Include: the need in our villages, what we will do, the results so far, the budget, partners, and how we measure success. Clear headings, convincing but factual.
• AI at Work · Block 4 · Practice file 4 · Project and figures are invented

Speaker notes: This is usually the most satisfying practice: in a few minutes they go from rough notes to three finished pieces – a proposal, a press release and a LinkedIn post. Encourage them to keep the conversation going: "make it more convincing for the alderman", "add a short risk section", "shorten it to half a page for the Supervisory Board". Check together: did Copilot keep the budget at €48,000 and does the breakdown still add up (32,000 + 6,000 + 7,000 + 3,000)? Did it invent partners or results that aren't in the notes? Copilot sometimes adds plausible-sounding numbers – a perfect lead-in to block 5.

Slide 31: Let's talk about it (debrief)

• Let's talk about it
• What surprised you?
• Where did Copilot get something wrong?
• Which task would you really use it for next week?
• AI at Work · Block 4 — Debrief

Speaker notes: Short conversation. The second question leads naturally into block 5: whatever mistakes they found (a wrong date, an invented promise, a generic tone) are exactly why we check everything. Note which task each person wants to use next week – that feeds into block 6.

Using AI safely and responsibly

Slide 32: Using AI safely (s5)

• Block 5
• Using AI safely
• You stay in charge – Copilot is the assistant

Speaker notes: Come back to the worries people raised in the icebreaker and address them here. Keep the tone practical, not scary: these are good habits, like checking a letter before it goes out on library letterhead.

Slide 33: Five golden rules (golden-rules)

• Five golden rules
• 1 You are the author. Read everything before you send or share it.
• 2 Check facts, names, dates and numbers – every time.
• 3 Protect people's data. No more personal details than the task needs.
• 4 Work tools for work. Use your work Copilot, not free public AI apps.
• 5 People decide about people. Hiring, reviews, absence stay human.
• AI at Work · Block 5 — Safe use · General good practice; follow your organisation's own rules where they exist

Speaker notes: The library chain has no AI policy yet, so these are general good-practice rules. Rule 3: patron data is protected by privacy law (GDPR / AVG). Avoid pasting in things like card numbers, addresses, health details or reading histories unless it's truly needed – and usually it isn't. Rule 4: Microsoft 365 Copilot with a work account keeps your data inside your organisation's Microsoft environment; free consumer AI apps may not, so work documents belong in the work tool. Rule 3 also covers staff data: absence records, performance reviews and salary details are sensitive. Rule 5: AI can help you structure a vacancy text or prepare a review conversation, but selecting candidates or judging a colleague is yours. Invite discussion: "Should Bibliotheek Westland write its own AI guidelines for staff? What would you put in them?" – a natural outcome for a management team (in the practice MT meeting, Marieke is asked to draft exactly that).

Slide 34: Spot the mistakes: find all four (act-spot-mistake)

• ACTIVITY
• Spot the mistakes: find all four
• Dear Ms Koppenol,
• Thank you for your email. We will send you the complete 2026 figures by 27 November. Visitor numbers in Maasdijk fell by 8%, so we have decided to move the Tuesday evening opening to Saturday morning. We look forward to seeing you on Tuesday 8 December at 10:00 at the town hall in Naaldwijk.
• Kind regards,
• Jeroen van der Berg, Manager Public Services
• Compare with the email thread from practice 1.
• AI at Work · Block 5 · Activity – an AI-style draft with planted errors

Speaker notes: Tell them this is the kind of draft Copilot could produce – it reads well, which is exactly the danger. Let them compare it with the email thread (practice file 1). The four mistakes: 1) "Complete 2026 figures by 27 November" – they agreed January–October plus a forecast by 27 November, and final figures by 31 January. 2) "Fell by 8%" – Jeroen wrote about 18%. 3) "We have decided to move the Tuesday evening opening" – not decided; the MT decides on 1 December, and this is exactly what they must not promise the municipality. 4) "10:00" – the director can only make the afternoon, so it should be 14:00. The date, place and signature are correct. Ask: "Which mistake would have caused the most trouble with the municipality and the council?" Usually the invented decision about Maasdijk – it could end up in council questions and the local press.

My first AI task, the 30-day plan and wrap-up

Slide 35: My first AI task (s6)

• Block 6
• My first AI task
• From practice to your own work

Speaker notes: Bring back the flip chart or sticky notes from the icebreaker with each person's "task I wish took less time".

Slide 36: Try it on your own task (my-task)

• ACTIVITY
• Try it on your own task
• 1
• Pick the task you named this morning
• 2
• Write a prompt with all four ingredients
• 3
• Run it, then improve it once
• 4
• Show the group your result
• Stuck? Look in the sample prompt library for a starting point.
• AI at Work · Block 6 · Worksheet in the participant handout

Speaker notes: This is the most important activity of the day: it connects the workshop to their real job. Use real but non-sensitive management work – e.g. a paragraph for the annual report, an agenda for the next MT, a summary of a policy document from the municipality or ProBiblio, a first draft of an annual plan section, an email to a partner organisation. Remind them of golden rule 3: no personal data about staff or members, so no absence cases or performance reviews in this exercise. Help each person one by one; with three people you can give everyone individual attention. Each person shows their prompt and result in about a minute. Praise good use of context and source.

Slide 37: Your 30-day practice plan (plan-30)

• Your 30-day practice plan
• Week 1
• Summarise one long email every day
• Week 2
• Draft one email or document with Copilot each day
• Week 3
• Try a meeting recap or an Excel question
• Week 4
• Share your best prompt with your team
• Small and daily beats big and once. Ten minutes a day is enough.
• AI at Work · Block 6 · Full plan in the follow-up pack

Speaker notes: Hand out the 30-day plan. Suggest they block ten minutes in their calendar each morning for the first week. Week 4 matters for managers: by sharing a good prompt with their team, they become the AI champions in their branch. Offer a short check-in (call or email) after two weeks if you can.

Slide 38: Thank you! (wrap-up)

• Thank you!
• Remember: AI is a well-read new assistant
• Ask well: Goal · Context · Source · Expectations
• Stay in charge: read and check before you send
• Please fill in the short feedback form · Questions later? fahriddinsr@gmail.com

Speaker notes: Recap the three big ideas, return to the hopes people shared at the start and check they were covered. Hand out the feedback form and give people a moment to fill it in before they leave. Tell them you'll send a follow-up email tomorrow with the prompt library and the 30-day plan.