Chapter 16 Exercises

These exercises focus on Gemini's Workspace integration, NotebookLM, and the specific workflows where Google's ecosystem provides genuine advantages. Exercises marked [Workspace] require a Google Workspace account with Gemini features enabled. Exercises marked [Advanced] work best with a Gemini Advanced subscription.

Section A: Gemini Interface and Web Grounding

Exercise 16.1 — Web Grounding Quality Comparison

Choose a topic where current information matters — a recent industry development, a regulatory change, a market trend. Send the same query to: - Gemini (gemini.google.com) - ChatGPT with browsing enabled

For each response, evaluate: - How current is the information? (check publication dates of cited sources) - How many distinct sources were consulted? - Were the sources high-quality and relevant? - Did either response present any information that appears inaccurate?

Document your findings. In which cases does Gemini's Google Search grounding provide an advantage? Are there cases where ChatGPT's browsing produces comparable or better results?

Exercise 16.2 — Multimodal Processing

Find any piece of video content relevant to your work — a recorded presentation, a conference talk, an industry briefing on YouTube. Share the YouTube URL with Gemini and ask:

  1. "Summarize the key points made in this video."
  2. "What is the speaker's central argument?"
  3. One specific question relevant to your work about the content.

Evaluate: how accurately does Gemini's summary match the video's actual content? Where does it excel at synthesis? Where does it miss important details?

Exercise 16.3 — Long Context Loading

Find three to five documents relevant to a current project or research question. Load all of them into a single Gemini conversation (use the file attachment feature for each document).

After loading: 1. Ask Gemini to describe each document briefly 2. Ask a synthesizing question that requires information from more than one document 3. Ask: "Where do these documents disagree or present conflicting information?"

Evaluate: how well did Gemini maintain context across all documents? Did the synthesizing question produce useful insights? What did the contradiction-finding surface?

Section B: Gmail Integration [Workspace]

Exercise 16.4 — Help Me Write for a Difficult Email

Identify an email you need to write that requires careful wording — a follow-up to an unanswered request, a negotiation message, a delicate feedback message, or a proposal follow-up.

Use Gemini's "Help me write" feature to generate a first draft. Then: 1. Edit the draft to match your actual voice and add any relational context 2. Compare the final version to what you would have written without AI assistance 3. Estimate the time saved and assess the quality difference

Note: for emails requiring genuine relationship sensitivity, plan to edit substantially. The AI draft handles structure; you handle relationship context.

Exercise 16.5 — Thread Summarization

Find a long email thread in your Gmail (20+ emails is ideal). Use Gemini's thread summarization feature.

Evaluate the summary: - Did it correctly identify the key decisions made? - Did it accurately capture open items or action items? - Was anything important missing from the summary? - Were any decisions mischaracterized?

This exercise calibrates your trust in email summarization for real professional use. Most users find it excellent for orientation but imperfect at capturing nuance.

Exercise 16.6 — Batch Draft Assistance

If you have a category of email you send repeatedly (customer follow-ups, meeting confirmations, weekly status updates, vendor inquiries), use Gemini to draft three to five of them in a single session.

After each draft, edit minimally and note what types of edits you consistently make. After the batch, write a "Help me write" prompt template that incorporates your consistent edits upfront — producing better first drafts for future sessions.

Section C: Google Docs Integration [Workspace]

Exercise 16.7 — Structure First, Content Second

For your next document that needs to be written (report, proposal, analysis, summary), start with Gemini rather than a blank page.

Describe to Gemini: what the document is for, who the audience is, what it needs to accomplish, and any constraints (length, format). Ask for a structural outline before any writing begins.

Review the outline through conversation — push back on structure that does not fit, add sections that are missing, remove sections that are unnecessary. Only after the structure is agreed should you begin writing content.

Evaluate: how much did the AI-assisted outline phase save versus your typical starting process? What did Gemini add structurally that you would not have thought of?

Exercise 16.8 — The Editing Pass

Take a document you have recently written (or are currently writing). Run two targeted editing passes with Gemini in the side panel:

Pass 1: "Edit the following text for concision only. Remove all unnecessary words without changing the meaning."

Pass 2: "Read this paragraph as a reader who does not already agree with my position. What would they find unconvincing or unclear?"

Apply your judgment to both sets of suggestions. What patterns do you notice in what Gemini catches versus what you catch in your own editing?

Exercise 16.9 — Cross-Document Synthesis in Docs

You have multiple source documents (research, reports, notes). Using the Gemini side panel in a new Docs document: 1. Describe your synthesis goal: what document are you trying to write? 2. Paste key passages from your sources as context 3. Ask Gemini to help draft synthesis sections that connect the source material

Evaluate the synthesis quality: does it correctly represent the sources? Does it add useful connective reasoning? Where does it require the most editing?

Section D: Google Sheets Integration [Workspace]

Exercise 16.10 — Formula Generation in Natural Language

Identify three formula challenges in your current work in Sheets — formulas you have been putting off because they are complex, or formulas where you are not confident in the correct syntax.

For each one, describe what you want to accomplish in plain English to Gemini in Sheets. Receive the formula, and verify it works correctly against your data.

Note the time saved versus your previous approach (documentation lookup, StackOverflow, trial and error). For most users, this is the highest-time-return Sheets feature.

Exercise 16.11 — Data Analysis with Gemini in Sheets

Upload or open a dataset in Google Sheets. Ask Gemini:

  1. "What patterns or trends do you observe in this data?"
  2. "What columns or data points appear anomalous or worth investigating?"
  3. "If I wanted to build a pivot table from this data that answers [specific business question], how would you structure it?"

Compare Gemini's analysis in Sheets to what you would get from uploading the same data to ChatGPT's Advanced Data Analysis. What are the strengths and limitations of each approach for your specific data work?

Exercise 16.12 — Synthetic Data Generation

Use Gemini in Sheets to generate a synthetic dataset for testing purposes. Ask it to generate 50 rows of sample data for a scenario relevant to your work (customer records, product performance, survey responses, financial transactions).

Review the generated data: - Is it realistic in format and value ranges? - Does it have sufficient variety for testing purposes? - Are there any unrealistic patterns that would invalidate tests?

Edit as needed. This exercise is valuable even if you do not currently work with Sheets data — synthetic data generation for testing is a broadly applicable technique.

Section E: Google Slides Integration [Workspace]

Exercise 16.13 — Presentation from a Prompt

Identify an upcoming presentation you need to prepare. Using Gemini in Slides (or the "Create a presentation about" feature), generate a first-draft structure from a prompt describing: - The presentation topic - The audience - The key points you want to make - The desired length (number of slides)

After receiving the generated structure: 1. Evaluate the slide organization and sequence 2. Identify which slides have useful content and which need major revision 3. Edit the content for accuracy, your specific voice, and any organizational nuance

Document: what percentage of the final presentation came directly from Gemini versus what required significant rewriting? Under what conditions would you use this feature more? Less?

Exercise 16.14 — Speaker Notes Generation

Take an existing presentation (one you use regularly) and ask Gemini to generate speaker notes for each slide. Provide the audience context and the purpose of the presentation.

Review the speaker notes: - Are they accurate elaborations of the slide content? - Do they add useful talking points not on the slide face? - Do they maintain a consistent speaking voice? - What would you need to change to actually use them?

This exercise is particularly useful for presentations you give to different audiences — Gemini can quickly adapt speaker notes for a technical versus non-technical audience.

Section F: NotebookLM

Exercise 16.15 — Your First NotebookLM Notebook

Set up a NotebookLM notebook for a current project or research topic. Upload at least five sources (documents, URLs, or a mix). After uploading:

  1. Ask NotebookLM to describe each source briefly
  2. Ask a synthesizing question: "What is the main argument or finding across these sources?"
  3. Ask for contradictions: "Where do these sources disagree?"
  4. Ask what is missing: "What important aspect of this topic is not covered by these sources?"

Evaluate: how does the source-grounded approach differ from asking the same questions in a general AI chat interface? What specific value did the citation feature provide?

Exercise 16.16 — Deep Research Query Sequence

Using a NotebookLM notebook with research sources for a real project, conduct a structured research session:

  1. Ask three specific research questions that require synthesizing information from multiple sources
  2. For each response, follow the citation links to verify the claims against the source text
  3. Ask one "what is not here" question: "What key question about this topic cannot be answered from these sources?"
  4. Request a briefing document: "Generate a concise briefing document on [specific topic] based solely on these sources, with citations."

Document what percentage of the briefing document's claims were accurately sourced and what required correction. This gives you a real calibration of NotebookLM's reliability for your domain.

Exercise 16.17 — Audio Overview Generation

Using an existing NotebookLM notebook (from Exercise 16.15 or 16.16), generate an Audio Overview.

Listen to the overview during a commute, walk, or exercise session. Evaluate: - Is it an accurate representation of the source content? - Does the conversational format make the content more accessible? - What would you use this feature for regularly? - What are its limitations for professional use?

Exercise 16.18 — Contradiction Detection at Scale

Find a topic in your work where you have multiple sources that may disagree — industry reports with different data, research papers with conflicting findings, internal documents with inconsistent policies.

Load all sources into NotebookLM and ask specifically: "Please identify any factual contradictions between these sources. For each contradiction, cite the specific passages from each source that conflict."

Verify the identified contradictions against the original sources. Is NotebookLM accurately identifying real contradictions, or is it finding surface-level differences that are not genuine contradictions?

Section G: Cross-Tool Comparison

Exercise 16.19 — The Same Task Across Three Tools

Choose a single task relevant to your work and complete it using: - Gemini (gemini.google.com) - ChatGPT - Claude

Good candidates: writing a two-page analysis of a current industry topic, reviewing a document for key issues, generating a structured plan for a project.

After completing all three, rate each output on: 1. Accuracy of content 2. Quality of reasoning 3. Appropriateness of format 4. Time to produce the output (including any editing you did)

This side-by-side comparison is the fastest way to develop personal calibration for which tool serves which task best.

Exercise 16.20 — The Workspace Displacement Test

For one full week, try to do all your AI-assisted work within Google Workspace's Gemini features (and NotebookLM), using the standalone AI chat interfaces only as a fallback.

After the week, reflect: - Which tasks were well-served by the in-product Workspace features? - Which tasks required you to go to a standalone interface? - What was the friction difference between using in-product Gemini versus opening a separate browser tab? - For the tasks that stayed in Workspace: did quality suffer, improve, or stay the same?

This exercise tests whether Google's integration advantage translates into real workflow improvement for your specific work.

Reflection Questions

  1. The Workspace integration advantage is real for certain tasks but not universal. Based on these exercises, for which specific tasks in your work does working inside Google apps (rather than a separate chat interface) provide genuine workflow value?

  2. NotebookLM's source-grounded approach is fundamentally different from general AI chat. In what professional contexts would this distinction matter most for you? When might you prefer a general chat interface's broader knowledge base over NotebookLM's source constraints?

  3. Gemini's quality is more variable than Claude's or ChatGPT's. Did you observe this in your exercises? How does variability affect your willingness to trust outputs for professional use?

  4. If your organization is standardized on Google Workspace, what is the single most impactful change you could recommend for how your team uses Gemini? What would need to be true (training, policy, subscription level) to implement it?