CloudspaceSESSION 01 / KNOWLEDGE & PRODUCTIVITY
Programme
LEARNING PATH & SLIDES
Built for the workshop.Use ← → to move between slides. Choose a slide from the learning path above.

Download exercise pack ↓
Session 01 / Learn · try · check
9:00–9:15

AI-Powered Knowledge & Productivity

NotebookLM / Gemini Notebook and GenAI for business value.

Who this is for

Business analysts, product managers, knowledge workers and non-technical stakeholders. No coding required.

Today’s outcome

Create a Business Briefing Kit from supplied documents: verified findings, an action list and an open question.

20–30 participants · Four hours, 9:00–1:00 · Six teams of 3–5 · A short project and presentation.
Session 01 / Learn · try · check
9:00–9:15

Start with a real work task.

What document task takes too much of your time?

  1. Choose one: summarise a meeting, compare proposals, review a contract, or research a market.
  2. Record confidence from 0–2 in the worksheet.
  3. Success today: create a notebook, manage sources, verify an answer, and produce a useful briefing.
Session 01 / Learn · try · check
9:00–1:00

Four hours: understand, practise, apply.

Each concept has a plain-language explanation, an example and an exercise.

TimeActivityEvidence
9:00–9:15Purpose and baselineOne work task
9:15–9:35GenAI, LLMs and groundingExplain limits
9:35–10:05Notebook workspace and sourcesFirst notebook
10:05–10:35Prompt, verify and saveChecked answer and note
10:35–10:50Break
10:50–11:10Three business use casesOne mini exercise
11:10–11:20Outputs and knowledge reuseReviewed output
11:20–11:35Best practice: RACI and Bloom’sOne checked framework response
11:35–12:25Business Briefing Kit projectBrief and peer check
12:25–12:55Six team presentations4 minutes + 1 minute feedback
12:55–1:00Exit checkIndividual reflection
Session 01 / Learn · try · check
9:15–9:35

What are GenAI and large language models?

Generative AI produces new content; an LLM generates language from learned patterns and the context you supply.

Example

Ask for a meeting summary: the model drafts a new summary rather than simply copying the minutes.

Why it matters

A fluent sentence can still be wrong. Your job is to supply relevant material and check facts, commitments and missing information.

Three-panel comic. Panel one: a worker carrying a huge stack of meeting papers labelled Notes, Action items and Decisions says, “Summarize this meeting.” Panel two: a robot at a laptop is fed a pile marked Transcript, Notes, Context and Jargon and replies, “I don’t photocopy. I predict the next useful words.” A label reads, “LLM equals language pattern engine.” Panel three: the worker and a colleague examine the robot’s meeting summary and she says, “Nice. Now check the facts, promises, and what got missed.” A caption reads, “GenAI creates new content. Confidence does not equal correctness.”
What happens when I ask generative AI for help?
Quick check: does a confident answer prove a claim is true? Explain why not.
Session 01 / Learn · try · check
9:15–9:35

What is NotebookLM?

An AI research assistant organised around notebooks and source materials.

What it does

Helps you ask questions across documents, summarise information and create reusable outputs.

Business value

Spend less time finding passages; spend more time checking meaning and deciding what to do. Measure time saved alongside corrections needed.

Example: combine meeting notes and a supplier proposal to prepare a source-backed manager briefing.
Session 01 / Learn · try · check
9:15–9:35

Grounding and RAG: bring evidence into the answer.

RAG means Retrieval-Augmented Generation: finding relevant material and using it as context for a generated answer.

  1. Start with source documents, such as a proposal and meeting notes.
  2. Relevant passages provide context for the question.
  3. The model generates an answer; citations let you inspect the supporting material.
This is a conceptual explanation, not a diagram of NotebookLM’s undisclosed internal implementation. Grounding reduces unsupported answers but does not guarantee accuracy.
Five-panel comic, "RAG: bring evidence into the answer — same AI, more context, less hallucination." One: a worker asks, "What are the approval steps for our new project?" Two: a robot says, "Let me search your documents," pointing to Project Proposal, Meeting Notes, Policy Guidelines and Past Projects. Three: the robot answers, "Based on your documents, here are the steps: prepare proposal, get department review, finance approval, final sign-off," marked with citations one, two and three. Four: the worker says, "Looks good... let me check the sources to be sure," and ticks all three sources. Five: both celebrate — "Now I know where it came from" and "Grounded answers make everyone happier." Footer: "RAG: same AI, just with your knowledge. Always check. Always learn."
How does grounding help?
Session 01 / Learn · try · check
9:35–10:05

Create your first notebook.

Use the familiar event pack before the business case.

  1. Open NotebookLM / Gemini Notebook and create a notebook named Community Learning Day.
  2. Download and unzip the exercise pack. Find morning-starter/01–03.
  3. Locate Sources, Chat and Studio. Reopen your notebook from the notebook list.
Session 01 / Learn · try · check
9:35–10:05

Three areas, three jobs.

Know where to go before trying advanced features.

Sources

Import and inspect the material you want to work with.

Chat

Ask questions, refine answers and open citations.

Studio

Create and review notes and available output formats.

Three-panel comic, "Three areas, three jobs — same notebook, different superpowers." One, Sources: import and inspect the material you want to work with; a worker gathers PDFs, slides, web links and notes into a folder marked My sources. Two, Chat: ask questions, refine answers and open citations; the worker asks, "Based on my sources, what are the key points?" and the assistant replies, "Here’s the answer with citations," showing citations one, two and three. Three, Studio: create and review notes and output formats; the worker says, "Turn this into a summary, study guide and FAQ," and those documents appear. Caption: from information to understanding.
Where do I start in a notebook?
Session 01 / Learn · try · check
9:35–10:05

Add, inspect and select sources.

An imported source and an AI-generated answer are different things.

  1. Add morning-starter/01-current-guide.md, 02-programme.md and 03-old-notice.md. If upload is unavailable, paste the text as a source.
  2. Open each source and check its title, contents and version.
  3. Select only the current guide and ask when registration opens. Repeat with only the old notice.
  4. Record why 10:00 in Room B is current and 9:00 in Room A is superseded.
Three-panel comic, "Add, inspect and select sources — different sources, different answers." One, Add your sources: a worker adds 01-current-guide.md, 02-programme.md and 03-old-notice.md to My sources. Two, Inspect each source: they check the title, contents and version; a panel shows 02-programme.md, Programme Guide, version 2.1, dated 15 August 2026 — "looks relevant." Three, Select the right source: "For registration questions, I’ll use the current guide," with 01-current-guide.md selected from the list. A note reads, "Right source equals better answer."
Why does source selection matter?
Session 01 / Learn · try · check
10:05–10:35

A useful prompt specifies the job.

Give a task, audience, source scope, output format and uncertainty instruction.

Try this prompt
Using only the current event guide and programme, write a three-bullet arrival briefing for a first-time attendee. Include time, room and what to bring. Cite each factual claim. If a detail is missing, say it is not documented.
Why this helps: the answer has a purpose and a checkable structure. Follow up with “Make it shorter without adding facts.”
Session 01 / Learn · try · check
10:05–10:35

A citation is a place to check.

Verify the passage supports the exact claim, including conditions and dates.

  1. Ask about registration, what to bring and parking fees.
  2. Open citations and compare each answer to the original passage.
  3. Record parking fees as not documented; do not fill the gap with a guess.
  4. Save one corrected, verified response as a note.
Three-panel comic, "A citation is a place to check — same answer, check the source." One, read the answer and find the citation: a worker reads an AI answer, "Parking at the venue is RM10 per day," carrying citation one; an arrow notes, "This is the citation — let’s check it." Two, open the source and verify: the worker says, "Hmm... let me check the source," and Source one, Venue Guide 2026.pdf, states "Parking: RM10 per day (weekdays only)," followed by "Weekend: RM15 per day. Last updated 12 August 2026." Three, use the verified information: the worker says, "Now I know! RM10 per day (weekdays only). I’ll note this down," and writes a note recording the fee, the weekend rate, the source and the date checked.
How do I turn a response into something I can rely on?
Session 01 / Learn · try · check
10:05–10:35

Practice: produce a checked arrival briefing.

Work individually, then swap with a partner.

  1. Write one prompt and improve it through a follow-up.
  2. Verify three facts against the current documents.
  3. Record one missing detail and save a note.
  4. Partner check: can someone else trace your facts back to the sources?
Checkpoint: mark each operation independent, prompted or retry needed in day1-worksheet.md.
Session 01 / Learn · try · check
10:35–10:50

Take fifteen.

Return ready to use the same method on business documents.

Next: contracts, meeting summaries and market research.
Session 01 / Learn · try · check
10:50–11:10

One method, three business applications.

Use the fictional business-starter pack. Create a separate notebook named Learning Hub Business Brief.

Contract analysis

Extract deliverables, payment stages and exclusions from a draft supplier contract. Separate proposed terms from agreed decisions.

Meeting summaries

Separate decisions, actions, owners, deadlines and unresolved questions. Do not turn a suggestion into an approval.

Market research

Compare the supplied options using stated criteria. Label missing evidence; supplied fictional notes are not current market research.

Session 01 / Learn · try · check
10:50–11:10

Example: proposed does not mean approved.

The supplier draft proposes RM12,000. The meeting records a RM10,000 budget cap.

Try this prompt
Compare the draft contract and meeting notes. List proposed terms, recorded decisions and conflicts in a table with citations. Has purchase approval been recorded? Do not infer approval from a proposal.
Expected: a RM2,000 mismatch and no recorded purchase approval. Ask the owner to resolve it. Practice: identify the contract’s training inclusion and exclusion.
Session 01 / Learn · try · check
10:50–11:10

Example: turn minutes into an action list.

A helpful summary preserves accountability.

Try this prompt
From the meeting notes, produce a table: action, owner, deadline, status and citation. Separate decisions from proposals. Write “not assigned” when no owner is given.
Expected: Aina requests revised pricing by 18 September; Kumar checks access by 19 September. A proposed pilot date remains tentative. Practice: find one unresolved issue.
Session 01 / Learn · try · check
10:50–11:10

Example: compare options without inventing evidence.

A recommendation should show its assumptions.

Try this prompt
Compare the two options in the research notes for a manager. Use columns for cost, support, missing evidence and source. Keep source facts separate from your recommendation. What must we verify before choosing?
Expected: Option A costs more than the budget cap; Option B support response time is unspecified. Practice: suggest one follow-up research question.
Session 01 / Learn · try · check
11:10–11:20

Save, generate and reuse knowledge.

An output is a draft to review, not a new authoritative source.

  1. Save a checked answer as a note and add your own decision question.
  2. Create one available Studio output, such as a report or audio overview. Review facts against sources. Use a written briefing if generation is unavailable.
  3. Copy or export a verified item. Inspect sharing permissions without inviting anyone or making it public.
Three-panel comic, "Save, generate and reuse knowledge — your notes today, better answers tomorrow." One, Save a checked finding: a worker says, "This looks correct. I’ll save it as a note with my question," beside a note card headed Parking fees listing RM10 per day (weekdays only), Source: Venue Guide v2.1, Checked 12 August 2026, with a Save to notebook button. Two, Generate useful outputs: "Now I can turn this into a summary, report or audio overview!" beside a Create in Studio menu offering Summary, Study guide, Audio overview, FAQ and Briefing doc. Three, Reuse and share safely: "Saved and verified. I can reuse this later or share it with my team!" with arrows to Share with team and Use in future notebooks. Caption: same knowledge, more value.
How do I reuse a finding I have already checked?
Use a notebook per topic, meaningful source names and version dates. Recheck imported content when source documents change.
Session 01 / Best practice
11:20–11:35

Name the framework. Describe the outcome.

Less explaining. A clearer starting structure. An output that is easier to review.

Shorter prompts

“Draft a RACI” names a familiar structure. You do not need to describe every column each time.

Describe the result

Say what you need: a responsibility table, a comparison, or a justified recommendation.

Easier to check and reuse

A consistent structure makes missing roles and evidence gaps easier to spot. Reuse the pattern with new sources.

Your questionFrameworkUseful output
Who does what?RACIResponsibility matrix with evidence and gaps.
How deeply should we think or learn?Revised Bloom’s taxonomyQuestions that move from recall to analysis, evaluation or creation.
Start short: source + framework or thinking task + outcome. Add detail only when the result needs it. Frameworks help structure an answer; they do not guarantee accuracy.
Three-panel comic, "Name the framework. Describe the outcome — a simple structure, clearer answers." One, Use a framework to structure the answer: a puzzled worker says, "Who does what? This is getting messy... I need a framework," beside RACI expanded as Responsible, Accountable, Consulted, Informed. Two, Draft the structured response: "Much clearer! Here’s the RACI for the project," beside a table headed Project: New Event, with rows Plan, Budget, Execute and Report and columns R, A, C and I, each cell naming a person. Three, A useful outcome: "Easy to review, share and reuse!" with ticks for clear roles, easy to check, reuse for new projects, and fewer gaps. Footer: frameworks turn messy questions into useful answers.
Why name a framework instead of writing a long prompt?
Session 01 / Best practice
11:20–11:35

RACI: clarify responsibility.

Make ownership and missing assignments visible.

RoleMeaningLearning Hub example
R · ResponsibleDoes the work.Aina requests revised pricing (BIZ-02).
A · AccountableOwns the outcome or final sign-off.Not documented in these notes.
C · ConsultedProvides input through discussion.Not documented; confirm with the owner.
I · InformedReceives progress or outcome updates.Not documented; confirm recipients.
Aim for one accountable owner per activity in an agreed RACI. In a source-based draft, leave missing authority “not documented” rather than inventing an assignment.

Reference: PMI · Responsibility assignment

Session 01 / Best practice
11:20–11:35

One sentence can describe the result.

RACI provides the structure; your prompt identifies the source and the required outcome.

Try this prompt
From BIZ-02, draft a RACI for the follow-up actions. Cite the roles and mark missing assignments “not documented”.
Expected: Aina is Responsible for pricing; Kumar for access. Migration responsibility and A/C/I roles are not established in this pack. Check the original passage before accepting a row.
Session 01 / Best practice
11:20–11:35

Bloom’s: ask for the thinking you need.

Select a useful level; every prompt does not need all six.

LevelQuestion using the business pack
RememberWhat is the budget cap?
UnderstandExplain the pricing conflict.
ApplyUse the cap to flag options needing revised pricing or approval.
AnalyseCompare costs, support and missing evidence.
EvaluateWhich next step is justified by the evidence and budget constraint?
CreateDraft a briefing separating findings, actions and recommendations.
Higher-level verbs do not guarantee better reasoning. Specify criteria and verify the result; a recommendation is not a recorded decision.

Reference: CDC · Revised Bloom’s taxonomy

Session 01 / Best practice
11:20–11:35

Try, verify, then apply it to the project.

No need to name every Bloom’s level. “Analyse” or “evaluate” can express the thinking you want.

Try this prompt
Analyse BIZ-01–03 against the RM10,000 budget. Recommend the next step, citing facts and flagging gaps.
  1. Run one RACI row or an Analyse/Evaluate prompt (3 minutes).
  2. Open the source passage; check a role, number or assumption and correct one weakness (2 minutes).
  3. Record the framework, why it fits and the check performed (2 minutes).
Keep it simple: name the sources, describe the outcome, then check the result. Add a short follow-up only where needed, such as “Separate confirmed roles from suggestions.” Open the prompt guide.
Session 01 / Learn · try · check
11:35–12:25

Project: Business Briefing Kit.

Help a manager decide what needs checking before a Learning Hub pilot.

  1. Use business-starter/01–03; produce a short executive summary with three verified findings and citations.
  2. Create an action list with owners and deadlines where documented. Use RACI to clarify roles, or a Bloom’s Analyse/Evaluate prompt to strengthen the briefing.
  3. Record one open question or evidence gap. Separate your recommendation from the recorded facts.
  4. Use materials/project-template.md. Name your framework, why it fits and one evidence check. Peer-check claims and revise before presenting.
Six teams of 3–5 cover 20–30 participants. Roles: curator, prompter, verifier and presenter; combine or share roles.
Session 01 / Learn · try · check
11:35–12:25

Build, check and rehearse.

Keep the output small enough for a busy manager.

TimeActivityEvidence
11:35–11:45Brief and source reviewAgree the decision question
11:45–12:10Create the kitSummary, actions and one gap
12:10–12:20Peer verificationCorrect one weakness
12:20–12:25RehearseFour-minute presentation
Session 01 / Learn · try · check
12:25–12:55

Show the evidence behind the briefing.

Four minutes to present; one minute of feedback per team.

  1. Explain the business question and show your summary.
  2. Open one citation and demonstrate what you verified.
  3. Show the action list, open question and one correction.
  4. Name your framework and explain how it improved the briefing or action list.
Assessment: correctness and evidence 35%; usefulness 30%; verification and limitations 20%; clear presentation 15%. Six teams × five minutes = 30 minutes.
Session 01 / Learn · try · check
12:55–1:00

What can you now do independently?

Complete the four exit questions and name your next real work task.

  1. Explain GenAI/LLM and one limitation.
  2. Explain how grounding and citation checking differ.
  3. Show one prompt improvement and its effect.
  4. Show a verified finding and the unresolved question in your kit.
Submit the kit and worksheet using the trainer’s agreed location. Include appropriate source access, never confidential real client documents.
Session 01 / Learn · try · check
Reference

Keep the working guides nearby.

Practice with safe sample data before using workplace material.

  1. Day 1 worksheet
  2. Business Briefing Kit template
  3. Exercise pack
  4. Google: answers and citations