AI Marketer (C2) — Exercises & Demos

Source of truth for all demos and exercises in this course. 12 modules, each with a demo (a worked demonstration the learner watches) and a hands-on exercise (learner self-assessed). Everything runs in Claude Code.

Internal doc — do not ship this README to learners. Everything else is learner-facing: the demo is watched, the exercise starter is the task, and the solution is shown after. Only this README is for the build team.

---

Folder structure (what's actually here)

Each module folder is content-named and unnumbered (see the rule at the bottom). Inside every module:

<module>/
├── demo/                          # worked demonstration (flat — no subfolders)
│   ├── walkthrough.md             # the demo worked through
│   └── example-output.md          # the finished artifact it produces
│                                  #   (data modules: CSVs + chart PNGs)
└── exercise/                      # hands-on, learner self-assessed
    ├── starter/
    │   ├── INSTRUCTIONS.md         # the TASK — what's expected (scenario, requirements, done-when)
    │   └── <assets>                # datasets / prompts / templates provided
    └── solution/
        ├── walkthrough.md          # the worked solution + Common Mistakes
        └── example-output.md       # a sample output (data modules: CSVs + charts)

What each file type is

File	Purpose
demo/walkthrough.md	The demo worked through — what the learner watches demonstrated
demo/example-output.md	The finished artifact the demo produces (data modules: CSVs + chart PNGs)
exercise/starter/INSTRUCTIONS.md	The task: scenario → What to produce → Requirements → Done when
exercise/solution/walkthrough.md	The worked solution (one strong example) + Common Mistakes
exercise/solution/example-output.md	A sample output — the clean deliverable (data modules: CSVs + charts)

All of the above is learner-facing (demos are watched; solutions are shown on the solution page). Only this README is internal.

Note on solutions: on the main branch, exercise/solution/ is an empty placeholder for the instructor to fill while recording. The generated-solutions branch contains generated worked solutions + demo outputs for review.

---

The 12 modules (folder → skill → brand)

#	Folder	Skill	Brand
SP1	goal-rich-briefing	Goal-Rich Marketing Briefing	Flowline (freemium SaaS)
SP2	stress-test-strategy	AI as Adversarial Reviewer	Flowline
SP3	evaluate-ai-tools	AI Marketing Tool Evaluation	generic (copywriting)
SP4	synthetic-persona-interviews	Synthetic Persona Design & Execution	Barkwell (dog beds)
SP5	synthetic-ab-test	Synthetic A/B Testing Discipline	Barkwell
SP6	scaled-audience-research	Scaled Synthetic Audience Research	Barkwell
SP7	cpa-cvr-forecast	AI-Augmented Forecast Construction	Steep (DTC tea)
SP8	validate-forecast	Forecast Validation & Calibration	Steep
SP9	ltv-to-bidding	AI-Augmented LTV Forecasting	Steep
M6	claude-code-knowledge-work	Marketing Knowledge-Work in Claude Code	Vessl (smart bottle)
M8	ai-generated-creative	AI-Generated Marketing Creative	Vessl
M9	background-agent-research	Single-User Agentic Delegation	Vessl

Brands repeat within a family (Flowline, Barkwell, Steep, Vessl) but no module references another — each is self-contained so learners can enter at any point.

---

What to review

• Demo walkthrough reads as a recordable ~8–12 min segment and lands the key takeaway.
• Exercise task is clear: a learner knows what to produce, the requirements, and when they're done — without being handed a step-by-step recipe.
• No spoilers in exercise starters (the answer isn't stated in the task).
• Solutions are correct and, for the data modules, match the provided datasets.
• Brand voice / scenario is appropriate for a global audience.
• Assets referenced by each INSTRUCTIONS.md are present in the same folder.

---

Open decisions (for Patrick / the author)

1. SP5 — the A/B demo needs a senior-dog persona panel; only the furniture-grade panel (for the exercise) exists. Build one, or seed live on camera?
2. SP9 — the cohort dataset uses real country codes (IN/CA/UK/US), engineered so India looks low-value until the confound is found. Keep, or switch to neutral region labels?
3. M9 — pin the exact background-agent surface to teach/record.
4. M8 — reframed from the dictionary's "Claude Design + Figma" to Claude Code + any image tool (Claude Design retired). Confirm.
5. M6 — now Claude Code (not "Claude Chat"); dictionary title should update.
6. SP7 — anchors on CPA, not the dictionary's "CPM" (agreed); dictionary wording to update.
7. Demo file model — demos currently carry brief + walkthrough + example-output; decide whether to keep all three or collapse to brief + output.

---

Building & standards

The Exercise Creation Resources folder holds Udacity's guidelines:

• Exercise Guidance.md — exercise design, instruction writing, solution requirements. Primary reference.
• Accessibility Standards.md — WCAG 2.1 AA.
• Real-World Content Guidelines.md — brands, trademarks, references (relevant to open decision #2).
• Third Party Images and Datasets.md — licensing & attribution.

⚠️ DO NOT NUMBER the exercises. Modular content may be reused across programs where order and count differ, so folder names are content-named, not numbered. (The SP#/M# labels above are for this build's reference only.)