Week 9 · Lecture outline

Week 9 — Lecture Outline · The AI Tool Landscape — Choosing the Right Tool

Using Artificial Intelligence · AI 101 Fall 2026 · Prof. Quinn Fictional sample

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective covered: Objective 3 — Work across modalities and choose the right AI tool for a given job from the current tool landscape.
SLOs touched: A (produce high-quality results through strong tool selection) · B (evaluate and use AI critically — including knowing when NOT to use it)
Meeting pattern: 2 sessions × 75 min = 150 min. Segment minutes below total ~150; scale to your own pattern.

Week at a Glance


The week's big question	"With dozens of AI tools available, how do I know which one to reach for — and when is no AI tool the right answer?"
By the end of the week, students can…	(1) map the major AI tool categories and name real tools in each; (2) match a given task to the best-fit tool and explain the reasoning; (3) distinguish general-purpose chatbots from specialized tools; (4) catch a mis-matched tool choice; (5) name tasks that should not be delegated to AI.
Key vocabulary	chatbot/assistant, image generator, audio synthesis, music generation, voice synthesis, video generation, research assistant, coding assistant, specialized vs. general-purpose tool, tool landscape, prompt injection (preview)
Materials	slides (Deck 9), the week's readings + official tool homepages, one approved assistant for the AI-critique moment, the tutorial
Timing note	8 segments, ~150 min total. Session 1 = Segments 1–4 (~75 min). Session 2 = Segments 5–8 (~75 min).

Segment 1 — Hook & the Promise (8 min) · Session 1 opens

Hook. Put a single realistic task on the slide — "I need to create a short jingle for my club's upcoming fundraiser event." Ask the class: "Which AI tool do you use?" Take quick answers. You'll likely get "ChatGPT" as the reflexive answer from most. Play it out: a chatbot can write lyrics, but it can't produce audio. A music-generation tool like Udio (https://udio.com) or Suno (https://suno.com) can produce the actual audio — but may not understand your detailed context as well as a chatbot. "The right answer depends on what exactly you need. That's this week."

The promise: "By Friday you'll be able to map the major AI tool categories, match the right tool to a specific job, and name at least three tasks you'd never hand to any AI — and you'll understand why that last list is just as important as the first."

Why it matters line: "The best tool for the job isn't always the one you already have open."

Segment 2 — The Tool Landscape: Six Categories, Real Examples (22 min)

Plain language first. The AI tool market has fragmented by modality and use case. A chatbot that writes beautifully is not the same kind of tool as an image generator or a music maker. Understanding the categories is more durable than memorizing any product list — the names change; the categories stay.

Walk through each category (keep a slide per category; name the real tools factually):

1. General-purpose chatbots / assistants
These are conversational AI tools that generate text in response to prompts. They're the Swiss Army knife of the landscape — good for writing, brainstorming, summarizing, Q&A, coding, planning, and much more.
- ChatGPT (OpenAI) — https://chatgpt.com
- Claude (Anthropic) — https://claude.com
- Gemini (Google) — https://gemini.google.com
- Copilot (Microsoft) — https://copilot.microsoft.com
- Grok (xAI) — https://grok.com
- Key insight: "All-purpose" means "very good at many things and not the best at any specialized one." A chatbot will write lyrics; it won't generate the audio.

2. Image generation tools
These generate visual images from text descriptions (prompts). Some allow editing and style control; their training data, style tendencies, and platform integrations differ.
- DALL·E (OpenAI, integrated into ChatGPT) — https://openai.com/dall-e-3
- Midjourney — https://www.midjourney.com
- Adobe Firefly (Adobe) — https://firefly.adobe.com
- Key insight: These are not chatbots — they take prompts and return images, not text conversations. Using one for a text task is a mismatch.

3. Audio / music / voice generation
Separate sub-categories: music generation (original compositions with vocals and instrumentation) and voice synthesis (converting text to a specific human-like voice).
- Udio (music generation) — https://udio.com
- Suno (music generation) — https://suno.com
- ElevenLabs (voice/audio synthesis) — https://elevenlabs.io
- Key insight: Udio and Suno make songs; ElevenLabs makes voices. These are distinct jobs. Don't conflate them.

4. Video generation
These generate video clips — motion, scenes, and sometimes audio — from text prompts or images.
- Sora (OpenAI) — https://sora.com
- Also in the category: Google Veo, Runway, Pika — mentioned for completeness but this week's focus is Sora as the flagship example.
- Key insight: Video generation is computationally intensive and outputs are currently short clips; best for specific visual content needs, not general communication.

5. Research assistants / notebook tools
These help you work with your own sources, documents, and data — not generating from scratch but analyzing, summarizing, and surfacing connections in material you provide.
- NotebookLM (Google) — https://notebooklm.google.com
- Also in the category: Perplexity, Deep Research features in ChatGPT/Claude/Gemini — mention briefly.
- Key insight: NotebookLM is grounded in your documents — it cites from what you upload, not from general training. This is a fundamentally different use case from a general chatbot.

6. Coding assistants
These specialize in writing, completing, explaining, and debugging code — integrated into editors or used as standalone tools.
- GitHub Copilot (Microsoft/GitHub) — https://github.com/features/copilot
- Cursor — https://cursor.com
- Claude Code (Anthropic) — https://claude.ai/code
- Key insight: A general chatbot can write code; a coding assistant is embedded in your workflow, sees your codebase, and provides context-aware suggestions. Different tool for different depth.

The landscape map summary: put all six categories on one slide with one real tool in each box. "These are six different kinds of tools built for six different kinds of jobs. The mistake is treating them as interchangeable."

Segment 3 — Matching the Tool to the Job (20 min)

The decision frame (teach this as a three-question test):
1. What is the primary output I need? (text? image? audio? video? code? an analysis of my documents?)
2. How specialized is the job? (general writing = chatbot; original music = Suno/Udio; grounding in my own documents = NotebookLM)
3. What are the failure modes? What would go wrong if I used the wrong category of tool here?

Three worked examples (do these live — show the decision, not just the answer):

Example A. Task: "Write captions for 10 Instagram posts for my club's spring event."
- Primary output: text → chatbot (ChatGPT, Claude, Gemini, etc.)
- Specialized? No — general writing with context.
- Wrong tool: image generator (won't write text captions; will generate images instead).
- Right pick: any of the approved chatbots.

Example B. Task: "Create original background music for a 90-second video."
- Primary output: audio/music → music generation tool (Suno or Udio)
- Specialized? Yes — need actual audio, not just lyrics.
- Wrong tool: asking a chatbot. It will give you lyrics or a description of music; it cannot produce audio.
- Right pick: Suno or Udio.

Example C. Task: "I have 40 pages of research notes. Help me find the three most recurring themes."
- Primary output: analysis of my specific documents → NotebookLM (or similar research assistant)
- Specialized? Yes — grounded in your own content.
- Wrong tool: a general chatbot that doesn't have your documents, or one where you'd need to paste all 40 pages and risk hitting the context window.
- Right pick: NotebookLM.

Common mis-matches to highlight:
- Using a chatbot to generate images (it can't — or does so poorly unless it has an image model integrated).
- Using an image generator to draft a speech (completely wrong category).
- Using Sora to generate a 90-minute film (wrong scale entirely).
- Using a general chatbot for code in a large codebase (a coding assistant with access to the repo context is better).

Memory hook: "Task first, tool second. What does the job need? Then which category fits? Then which specific tool?"

Segment 4 — Misconceptions + Quick Interaction (18 min) · Session 1 closes (~75)

Name the misconceptions out loud, then cure each:

❌ "One chatbot does everything equally well."
✅ Cure: chatbots are versatile but they're one category. An image generator, a music tool, and a research-notebook tool each do jobs a chatbot can't. Even within chatbots, strengths vary.
❌ "Image generators and chatbots are the same kind of tool."
✅ Cure: different architectures, different outputs. An image generator takes a text prompt and returns pixels, not a conversation. They are not interchangeable.
❌ "The newest tool is always the best tool for every job."
✅ Cure: recency ≠ fit. A three-year-old specialized tool built exactly for your job will usually outperform a brand-new general-purpose tool that does your job as a side feature.
❌ "I just need one tool and I'm set for life."
✅ Cure: the landscape changes — tools are released, updated, and sometimes discontinued. The durable skill is knowing how to evaluate a tool for a job, not which specific product is "best" this month.

Interaction — Classification Round (rapid-fire, ~10 min):
Put 8 tasks on a slide. Students call out the best-fit tool category (not product name), then one classmate names a specific tool in that category. Move fast: "Original voiceover for a podcast episode" → audio/voice synthesis → ElevenLabs. "A Python function that parses a CSV file" → coding assistant → GitHub Copilot or Cursor. "Three visual concepts for a logo" → image generation → Midjourney or Firefly. "Summary of a 30-page PDF you uploaded" → research assistant → NotebookLM…"

Segment 5 — When NOT to Use AI (15 min) · Session 2 opens

Hook back in: "Last session: which tool to use. Today: when to use none of them — and how to stay up to date as the landscape shifts."

Plain language first. Every tool in the landscape is good at some things and genuinely bad at others. But beyond "bad at it," there are tasks where delegating to AI — no matter how capable the tool — is the wrong call for reasons of ethics, safety, accountability, or skill development.

Category 1 — Tasks where errors are consequential and unverifiable in real time.
Medical triage, legal advice, emergency decision-making. AI can generate plausible-sounding responses, but a wrong answer causes real harm, and most users can't catch the error fast enough. "AI can support research; it cannot substitute for professional judgment where the stakes are immediate."

Category 2 — Tasks where the human doing the work is the point.
Learning a skill, developing your writing voice, building a relationship, having an honest conversation. If AI writes your personal statement, you haven't practiced writing; if AI conducts your networking call, you haven't built the relationship. The shortcut undermines the goal.

Category 3 — Tasks where the output becomes your credential or legal responsibility.
Signing your name to AI-generated work as if it's your own analysis in a professional or academic context (without disclosure) raises integrity and accountability issues. Full unit on this in Week 15; mention briefly here.

Category 4 — Deeply personal decisions.
Major life choices — career pivots, relationship decisions, health plans — can use AI to research options, but the decision itself should be yours. AI can model scenarios; it cannot weigh what matters to you.

Memory hook: "Before you delegate: who's accountable if it's wrong? And what do you lose by not doing it yourself?"

Segment 6 — Live Demo: Tool Bake-Off (18 min)

Set it up: "This is the exact move you'll practice in Studio 9 — give the same task to two tools and compare."

Do it live (narrate every step):
1. Pick a task: "Generate a short, upbeat caption (under 50 words) for a community event flyer."
2. Run it in ChatGPT — show the result.
3. Run the same prompt in Claude — show the result.
4. Compare live: which is better? For what criterion? (Tone? Specificity? Brevity?) Point made: same prompt, slightly different strengths — knowing this comes from trying both.
5. Now try the wrong category: paste the same caption prompt into an image generator's interface. "What happens? It generates an image, not a caption. The mismatch is immediate and obvious — and that's the lesson."
6. The verify beat: "For Studio 9, you'll also need to catch where one of the tools fails or fabricates — what specific claim in the output would you check?"

Land the key idea: tool selection is a learnable judgment call. The Studio gives you a real chance to practice it.

Misconception + cure:
- ❌ "Comparing tools takes too long — I'll just use the one I know."
✅ Cure: the comparison itself is often 5 minutes and teaches you more about both tools than an hour of tutorials. Do it once and you'll know when each is better.

Segment 7 — Staying Informed: The Tool Landscape Changes (18 min)

The problem: if you memorize a product list, it's out of date in 6 months. Tools get updated, acquired, discontinued, or replaced.

The durable habits (teach these as a short list):
1. Follow official product channels. Every major tool has an official release-notes page or blog. Bookmark a few — not review sites, which lag behind.
2. Test a new tool on a real task. The fastest way to understand a new tool's strengths is to use it on something you actually care about. Five minutes of real use beats 20 minutes of reading.
3. Ask the tool itself. A chatbot (or the tool's help docs) can often tell you what it does best and what it can't do yet. That's a reasonable starting point — with the caveat that self-descriptions can be optimistic.
4. Trust the task, not the hype. Announcements and demos show best-case scenarios. Your job is to find the edge cases: "Where does this break? What does 'impressive demo' hide?"
5. Know when a 'new' thing is actually new. Some features get rebranded; some are genuinely transformative. The context-window / agent / multimodal expansions were real step changes; a UI update is not.

The course's promise: "By the end of this term you'll have the judgment to evaluate any new AI tool that appears — because you understand the categories and you have the prompting and verification skills to put one through its paces."

Segment 8 — Technology Workflow + AI-Critique, Callback & Hand-off (12 min) · Session 2 closes (~75)

Technology workflow — the tool-selection process:
1. State the job — what output do I need? (text / image / audio / video / code / document analysis)
2. Identify the category — which category of AI tool is built for this?
3. Pick a specific tool — from that category, which is accessible, current, and appropriate?
4. Try it — run the task, assess the output.
5. Catch the mis-fit — if the result is wrong in kind (not just quality), you may have the wrong tool, not a bad prompt.

AI-critique moment (students verify, not consume):

Paste this into a chatbot: "List the best AI image generator for creating photorealistic product shots. Include the specific resolution each supports and their monthly pricing." Notice what happens: the chatbot will confidently answer, but specific resolutions and pricing change frequently and are the exact kind of detail AI fabricates or gets out of date on. Your job: flag every specific number or price claim as needing live verification. For tool capability claims specifically, the official product homepage is the only reliable source. This is the AI-critique skill you'll practice in Studio 9.

Callback + tease:
- Callback: "Week 7 we built the multimodal skills — voice, audio, images, documents. This week we zoomed out: the whole landscape, and how to navigate it. The two weeks belong together."
- Tease next week: "Next week we go deep on hallucination and verification (Objective 4) — the shapes hallucinations take, why they happen, and the full verification workflow. You've been practicing this all term; Week 10 is where we make it systematic."

Hand-off (the week's graded work):
- Lecture Tutorial 9 (AI tutor, share-link submission) — tool categories, matching, and "when not to use AI."
- Quiz 9 (no AI), Discussion 9 ("master one or juggle many" + "what should never be handed to AI"), and Assignment 9 (tool→job matching; choose-and-justify scenarios).
- AI Build Studio 9 — "Tool Bake-Off / Tool Map" — compare two or three tools on the same task, OR build your personal tool map.

Instructor FAQ — Common Stumbles

Student says / does	Quick cure
"Can't I just use ChatGPT for everything?"	ChatGPT is great for text; it can't generate audio, and it's not optimized for your own documents the way NotebookLM is. Using it for everything is a choice, but it means missing what specialized tools do better.
Confuses Suno/Udio with ElevenLabs.	Suno/Udio make music (songs with instruments + vocals); ElevenLabs makes voices (text-to-speech, voice cloning). Different output, different job.
Thinks DALL·E is the same as ChatGPT.	DALL·E is an image-generation model; ChatGPT is a conversational text model. Some ChatGPT plans include image generation via DALL·E — but they're separate capabilities, not the same thing.
"Isn't Grok just the same as ChatGPT?"	Grok (xAI) is a separate chatbot with its own model. Chatbots in the same category still have different strengths, training approaches, and integrations.
Claims NotebookLM is a chatbot.	NotebookLM is a research assistant grounded in your documents. It won't answer from general training the way a chatbot does — it cites from what you uploaded.
"The newest tool is always the best."	Recency ≠ fit. Evaluate for the specific job, not the launch date.

Scope flag

This outline stays within Objective 3 at the tool-landscape level (categories, tool→job matching, mis-matches, knowing when not to use AI, staying informed). The deeper verification and hallucination mechanics are Week 10. Real products named factually using official homepages as the authoritative source; the instructor and institution are fictional. Product-accuracy gate: PASS — every tool named is a real, current product; links point to official pages; no version numbers, pricing, or feature specifics are asserted (those require live verification).

~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com