Foundations Review

A cumulative review of Chapters 01-03 (The Org Age of AI, Claude in Your Pocket, Making Things with Artifacts). Ten questions, drawn from across the chapters and deliberately mixed together — interleaved practice is harder than chapter-by-chapter review, and that difficulty is exactly what builds retention.

How to use this review:

Pass bar: 9 of 10. Higher than lesson quizzes, because this is where knowledge proves it stuck.
Answer in writing, in your own words, before checking anything. Constructing the answer is the exercise.
With a tutor: it will run the review one question at a time, grade against the key, and generate fresh variants for any retake.
If you miss the bar — or barely clear it — come back tomorrow, not in five minutes. Spaced retakes are dramatically better for retention than immediate ones. The gap is a feature.

Check your understanding

Q1. Your business partner asks Claude a strategic question, gets a generic answer, and shrugs “it doesn’t know our business.” Describe the two-step response: what fixes part of the problem in the next five minutes, and what category of gap remains no matter how good the prompt gets?

Q2. You have 20 minutes of dead time on a train, and a process in your head that has never been written down. Which capture technique do you reach for, what exactly do you say to start it, and why does it beat opening a blank document tonight?

Q3. Sort these onto the maturity ladder and justify each in one line: (a) a bakery where all supplier rules live in the owner’s head, (b) a consultancy where every partner has their own elaborate private prompt library, (c) an agency with a shared, documented playbook that their AI tools follow.

Q4. A colleague photographs a whiteboard, gets a perfect transcription, and says “neat trick.” Explain why camera capture is strategically more than a trick — what class of organizational information does it unlock, and what would otherwise happen to that information?

Q5. Design the artifact: your team keeps asking you the same fifteen how-do-I questions. Specify who uses it, how they interact with it, where it runs — and then name the two iteration feedback sources you will use before calling it done.

Q6. The staff cheat-sheet artifact is now relied on daily. New vendor arrives. Walk the correct update sequence, name the principle behind it, and describe the failure that hits teams who do it the other way around.

Q7. Which goes where and why: the quarterly numbers your accountant files, the searchable policy reference your support team uses hourly, the org chart a new hire skims once?

Q8. You start refining a captured document on your phone in a waiting room, continue at your desk, and consult the result on-site the next day. Name the property that makes this seamless, and one deliberate habit that exploits it.

Q9. “Everyone here already uses ChatGPT, so we’re ahead.” Using the transitions framework, explain what this organization has and has not achieved, and the specific work its next step requires.

Q10. A published artifact link leaks outside the company. For each of these page types, is that an incident or a shrug: the vendor code cheat-sheet, the customer escalation list with names and numbers, the margin calculator? State the rule that decides.

Answer key — attempt every question first

Answer key

Q1

Model answer: Five-minute fix: re-ask with context — attach the relevant documents, state constraints, define what a good answer looks like. What remains: organizational knowledge that is undocumented (tacit rules, exceptions) or lives in systems (running totals, records) — the legibility gap that prompting cannot close.

Pass criteria: both steps present: context-craft fix AND the persistent gap; gap characterized as undocumented/system-resident, not ‘AI limitations’

Q2

Model answer: Voice-mode interview: “Interview me about [process]. One question at a time, dig into edge cases like you’re writing our operations manual.” Answering beats authoring — questions extract expertise that is invisible to its owner, and a blank page extracts nothing.

Pass criteria: voice interview chosen; prompt includes one-at-a-time + edge cases; answering-vs-authoring rationale

Q3

Model answer: (a) L0 Tribal — knowledge in one head, nothing written; (b) L1 Experimenting — individual AI use, nothing shared or compounding; (c) L2 Legible — work described in shareable form machines follow. Justifications must reference sharing/documentation, not tool sophistication.

Pass criteria: all three placed correctly; justifications turn on shared/explicit knowledge

Q4

Model answer: It unlocks the paper layer — invoices, delivery notes, whiteboards, handwriting — where a large share of operational truth lives. Without capture it stays untyped and unsearchable, permanently outside every digital workflow, because retyping was never worth ten minutes; a photo makes it worth ten seconds.

Pass criteria: identifies physical/paper information as a class; names its default fate (never digitized) and the economics change

Q5

Model answer: Any coherent spec: users (the team), interaction (searchable FAQ), platform (their phones/laptops). The two feedback sources: the real device in real context, and watching a real user stumble — then reporting observations verbatim to Claude.

Pass criteria: spec covers who/how/where; names real-device AND real-user feedback

Q6

Model answer: Update the source-of-truth document (the rules/skill) first, then regenerate and republish the page. Principle: the page is a view, not the home. Backwards teams edit the page directly: truth forks, and the next regeneration silently erases their page-only edits.

Pass criteria: sequence correct; view-not-home principle named; fork/overwrite failure described

Q7

Model answer: Accountant: spreadsheet — a filing workflow wants files. Support team: artifact — high-frequency interactive lookup wants a tool at a stable link. Org chart: a document is fine — read once, no interactivity needed. Rule: the consumer’s usage pattern picks the format.

Pass criteria: all three assigned sensibly; consumer-determines-format rule stated

Q8

Model answer: The account is the workspace — every device is a window onto the same conversations, so work travels without export. Habit example: capture rough on the phone the moment things occur to you, refine at the desk, use in the field — or keep one running conversation per active piece of work.

Pass criteria: account-level continuity named; a deliberate exploiting habit given

Q9

Model answer: They have reached L1: individual experimentation. They have NOT made the organization legible — nothing shared, nothing a machine could follow, knowledge still tribal. Next: T1 — capture and structure how the work actually happens (dictation, documents, then skills) so it is shareable.

Pass criteria: L1 diagnosis; names T1/legibility as next; concrete capture-and-share work described

Q10

Model answer: Cheat-sheet: shrug — internal reference data with no sensitive content. Escalation list with personal details: incident — PII on an uncontrolled public link. Calculator: shrug — logic, not data. Rule: publish only what would be acceptable on the open web, because a published link is readable by anyone who has it.

Pass criteria: three correct verdicts; anyone-with-the-link rule stated