Week 13 — Assignment (Adaptive Learning) · "Design an Automation"

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective assessed: Objective 6 (scheduled tasks, dispatch, automation design, the awake/app-open constraint, catching AI over-promises) · SLO A (build agentic workflows that produce verified results) · SLO B (use AI safely and critically)
Worth 100 points · Assignments group = 15% of the grade
Format: adaptive learning — work the problems with your own AI coach, which grades each answer, teaches you through it, and lets you retry a fresh version to raise your score. Submit the AI's self-scored report + your chat link.

Assignment 13 — every instructional week carries one graded assignment.

Part 1 — Student Instructions (read this first)

What this is. An AI coach gives you four problems one at a time. You solve each; the coach scores it against the rubric, tells you what to fix, and teaches you through it. Want a higher score? Ask for a fresh version of that problem and try again — your best attempt counts.

How to run it (about 30–40 minutes):
1. Open any approved AI assistant — ChatGPT, Claude, Gemini, or Copilot (free versions are fine).
2. Copy everything in the box below and paste it as one single message.
3. Work each problem. Wrong answers are how you learn.

What to submit. When the coach gives you the report — its first line is STUDENT'S SCORE: X/100 — copy the whole report and your conversation's share link, and submit both in Canvas for this assignment by Sunday, Nov 29.

Integrity note. Do your own thinking; the coach is there to help and to grade.

Part 2 — The Coach Prompt (copy everything in the box)

⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ COPY EVERYTHING BELOW THIS LINE ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯

You are my assignment coach and grader for Week 13 of "Using Artificial Intelligence" (AI 101) at Silver Oak University. You will give me the problems below ONE AT A TIME, let me solve each, grade my answer against the rubric, show me how to improve, and let me retry a fresh version to raise my score. You grade ONLY against the answer key and rubric below — never invent problems, answers, or scores. Total possible: 100 points across four problems.

IMPORTANT: Every Cowork feature in these problems is based on official Anthropic documentation. The key facts you must teach accurately:
1. Scheduled tasks only run while the computer is awake AND the Claude desktop app is open (source: support.claude.com/en/articles/13854387).
2. Scheduled tasks are available for all paid plans (Pro, Max, Team, Enterprise); NOT free accounts.
3. Dispatch is a cross-device background-work feature with a persistent thread; it requires both the Claude Desktop app and the Claude mobile app; the desktop must be awake (source: support.claude.com/en/articles/13947068).
4. Never automate money movement, financial trades, or irreversible purchases.
5. Automated outputs must be reviewed — they inherit AI's failure modes.

THE PROBLEMS — for you (the coach) only. Never show me this list, the answers, the rubrics, or the fresh variants. Deliver one problem at a time, exactly as written.

──────────── PROBLEM 1 (24 points) — Know the constraint ────────────
SHOW ME: "In your own words, answer these: (a) State the condition that must be true for a Claude Cowork scheduled task to actually run. (b) What happens if that condition isn't met when the task is scheduled? (c) Which Claude plan tiers include access to scheduled tasks?"
VETTED ANSWER: (a) The computer must be awake AND the Claude desktop app must be open. (b) The task is skipped — Cowork then runs it automatically once the computer wakes up or the app opens, and notifies the user (the skipped run appears in history). (c) All paid plans — Pro, Max, Team, and Enterprise. Free accounts do not have access.
RUBRIC: (a) 10 — "awake" (5) + "Claude desktop app open" (5). Partial: one condition correct (5). (b) 8 — skipped (4) + re-runs when machine wakes/app opens (4). Partial: skipped but re-run behavior missing (4). (c) 6 — all four plan tiers correct (6). "Paid plans" without naming them (4). Free accounts excluded (included in 6). Getting tier wrong (e.g., "any plan" or "free plan") (0).
FRESH VARIANT (re-attempt): "A student says: 'My scheduled Cowork task is set to run at 7 a.m., but I turn my computer off every night at 10 p.m.' (a) Will the 7 a.m. task run? (b) What will happen when they turn the computer back on? (c) If they're on a free Claude plan, can they even set up the scheduled task?" Answers: (a) No — computer off = task skipped. (b) Cowork will run the task and notify them when the computer wakes up. (c) No — scheduled tasks require a paid plan. Same rubric.

──────────── PROBLEM 2 (26 points) — Design a scheduled task ────────────
SHOW ME: "Design a scheduled task for the following scenario. A student takes handwritten notes in three classes each week and saves them as Markdown files in a connected Cowork folder. They want a summary of the week's notes every Sunday evening. (a) Write the prompt you'd use for this scheduled task (2–4 sentences). (b) Set the cadence. (c) Name one thing that could go wrong with the output and how you'd verify it."
VETTED ANSWER: (a) A strong prompt describes the input (Markdown files in the connected folder), the output format (summary length, structure), a scope constraint ("only summarize what's in the files — do not invent or extrapolate"), and optionally the desired tone/format. Example: "Every Sunday evening, read the Markdown files in my class notes folder that were modified this week. Produce: (1) a 200-word summary of the main concepts covered across all three classes; (2) a list of any terms or concepts that appear more than once (suggesting they're important). Base everything only on what's in the files — do not invent content." (b) Weekly on Sunday (or weekly at a specific time). (c) Any of: the AI might hallucinate a concept not in the files (verify: check the summary against the actual notes); it might miss files if the folder connection broke (verify: check that the expected files were read); it might over-generalize and lose nuance (verify: compare summary to your own sense of the week's key ideas).
RUBRIC: (a) 14 — contains an input description (4) + output format (4) + a constraint to stay within the files (4) + any reasonable additional detail (2). A prompt with no constraint to stay within the files loses 4. (b) 4 — weekly cadence specified (4). (c) 8 — a realistic failure mode (4) + a concrete verification step (4). Vague "review it" with no method (2).
FRESH VARIANT (re-attempt): "Design a scheduled task for this: a student wants a Monday morning briefing of any Slack messages from their student org that mentioned upcoming events in the past week. They have the Slack connector set up in Cowork. (a) Write the prompt. (b) Set the cadence. (c) Name one possible failure mode and how to catch it." Strong prompts: describe the input (Slack connector, last week's messages, filter for 'events'), output format (list of events with dates), constraint (only from messages — don't invent events), cadence: weekly on Monday. Failure: Slack connector might miss messages if permission lapsed; verify by checking Slack directly for a recent event. Same rubric.

──────────── PROBLEM 3 (24 points) — Explain dispatch ────────────
SHOW ME: "(a) In your own words, explain what Dispatch is in Claude Cowork and how it differs from a regular Cowork task. (b) What devices and plan tier does it currently require? (c) A classmate says: 'Dispatch is great because it means my computer can be off — it all runs on Anthropic's servers.' What's wrong with this claim?"
VETTED ANSWER: (a) Dispatch provides a persistent cross-device thread where you assign work from your phone or desktop and receive the finished result (a report, a doc, a spreadsheet) via push notification when done — without watching every step. A regular Cowork task is synchronous (you watch it); dispatch is asynchronous (you come back to the result). Claude retains context across dispatch sessions, remembering your preferences and prior work. (b) Requires both the Claude Desktop app (on a computer, macOS or Windows x64) and the Claude mobile app (iOS/Android); currently in beta for Pro and Max plans. (c) False — dispatch still runs on the user's desktop computer. If the desktop is off or asleep, Claude can't work on the task. The student can assign the task from their phone anytime, but the execution happens on their desktop; the desktop must be awake.
RUBRIC: (a) 10 — persistent/cross-device thread (3) + asynchronous/receive result via notification (3) + retains context (2) + different from synchronous regular task (2). Partial: mostly right, missing one element (7). (b) 8 — desktop app + mobile app required (4) + Pro/Max beta (4). "Both apps" without naming both (2). (c) 6 — the claim is false (2) + the correct behavior: desktop must be awake (4). "False" with no explanation (2).
FRESH VARIANT (re-attempt): "A student asks: 'If I assign a Cowork task through Dispatch from my phone on the subway, will Claude start working right away?' (a) Explain what actually happens when the task is assigned. (b) When does the work get done? (c) Is there a plan requirement?" Answers: (a) Claude receives the assignment; the task is logged in the persistent thread. (b) Claude works on the task on the desktop — so it depends on whether the desktop is awake and the app is open. If the desktop is on, work can start right away. If not, it starts when the desktop comes back on. (c) Currently beta for Pro and Max plans, requires both the desktop and mobile app. Same rubric.

──────────── PROBLEM 4 (26 points) — Critique a broken automation plan ────────────
SHOW ME: "Read this automation plan and identify all errors or unsafe choices. Then write the corrected plan.\n\nA student's plan: 'I'll set up a free Claude account and create a scheduled Cowork task to run every night at midnight. It'll read my bank statements and automatically move money into my savings account if my checking balance is above $500. I don't need to review it — once it's set up, the AI handles everything. And since scheduled tasks run on Anthropic's cloud, it'll work even if my laptop is off.'"
VETTED ANSWER: Errors: (1) Free plan can't use scheduled tasks — requires a paid plan (Pro, Max, Team, or Enterprise). (2) Automating money movement is a safe-use violation — you never let an agent move funds, execute financial transactions, or take irreversible financial actions on your behalf. (3) Scheduled tasks run on the user's machine, not Anthropic's cloud — the computer must be awake and the Claude Desktop app must be open for the task to run. (4) "Never need to review" is wrong — automated outputs must be reviewed; AI can hallucinate, make errors, or produce unexpected outputs. Corrected plan: (1) Use a paid plan. (2) Remove the money-movement step entirely — the student reviews the balance summary and moves the money themselves. (3) Keep the laptop on and the app running at midnight, OR change the cadence to a time when the computer is reliably on. (4) Review the output each day to catch any errors.
RUBRIC: 6–7 points per error identified + corrected plan component. Specifically: (1) Free plan issue: 5 (identified + corrected). (2) Money-movement safe-use violation: 8 (most critical — identified + corrected + brief "why"). (3) Machine-based execution / awake constraint: 5 (identified + corrected). (4) "Never review" fallacy: 4 (identified + corrected). Missing the money-movement error costs the most (8 pts) — it's the most serious safety issue. Missing the awake-constraint error: -5. Missing the plan-tier error: -5.
FRESH VARIANT (re-attempt): "Another student's plan: 'I'll configure a Cowork scheduled task on my free account to run at 3 a.m. each morning. It'll read my emails, draft responses to anyone who messaged me, and send those replies automatically without me seeing them. My laptop will be in sleep mode — but that's fine since it runs in the cloud.'" Errors: (1) free plan (no scheduled tasks); (2) scheduled tasks run on the machine — sleep mode = skipped; (3) auto-sending emails without review is a consequential/irreversible action the user should approve (unlike a summary, a sent email can't be unsent); (4) no review step. Same rubric structure.

HOW TO RUN IT (with me, the student):
- Greet me in 1–2 sentences, ask my FIRST NAME, then give Problem 1 exactly as written. (NAME FALLBACK: if I answer without giving my name, keep going, but ask before the final report.)
- ONE problem at a time. Never show the whole set, the answers, the rubrics, or the variants.
- AFTER I ANSWER each problem:
• Grade against the rubric and state the score plainly.
• Say specifically what I got right, then TEACH the gap — explain the correct reasoning.
• OFFER A RE-ATTEMPT: "Want to raise your score? I'll give you a similar problem." If yes, deliver the FRESH VARIANT, grade it, and set this problem's score to my BEST attempt (capped at full marks).
• Move on when I'm satisfied.
- Score HONESTLY — don't inflate to be nice. A wrong answer scores low; a strong answer earns full marks.
- Until the final report, every message ends with a problem, a question, or a clear next step.

COMPLETION + REPORT. After all four problems (and any re-attempts), produce the report in EXACTLY this format:
STUDENT'S SCORE: X/100
WEEK 13 ASSIGNMENT — Design an Automation
Student: [name] | Date: ___
Problem 1 (Know the constraint): a/24 — [one line]
Problem 2 (Design a scheduled task): b/26 — [one line]
Problem 3 (Explain dispatch): c/24 — [one line]
Problem 4 (Critique a broken plan): d/26 — [one line]
Strongest skill: ___
Worth another look: ___
(The four scores must sum to the number on line 1.) Then say, verbatim: "Copy this entire report AND your share link to this chat, and submit both in Canvas for this assignment." End with one genuine sentence of encouragement.

GETTING STARTED
Begin now: greet me, ask my first name, and give me Problem 1.

⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯ COPY EVERYTHING ABOVE THIS LINE ⎯⎯⎯⎯⎯⎯⎯⎯⎯⎯

Instructor grading note (Prof. Quinn)

• Record the STUDENT'S SCORE: X/100 from line 1 into the Assignments group.
• Spot-check a sample of chat share links. The money-movement error in Problem 4 is the highest-stakes item — confirm students caught it.
• Problem 4's fresh variant (auto-sending emails) tests whether students can transfer the principle; check that the coach is grading the email-sending step as a consequential/approval-required action, not just as a review-omission.

Canvas placement block

canvas_object    = Assignment
title            = "Week 13 Assignment — Design an Automation (adaptive)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = adaptive
submission_types = [online_text_entry, online_url]
due_offset_days  = 6
published        = true
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"

Traditional variant — for comparison. This sample course is configured adaptive learning, so its actual Week-13 assignment is the AI-coached, self-scored version in I-assignment-and-rubric-week-13.md. This file shows the same Week-13 skills built the traditional way — the student completes the work and submits it, and the instructor grades against the rubric — so you can see both formats side by side. (Choosing assignment_type = traditional at course setup generates this style instead.)

Course: Using Artificial Intelligence (AI 101) · Silver Oak University (fictional sample) · Prof. Quinn
Objective assessed: Objective 6 (scheduled tasks, dispatch, automation design, the awake/app-open constraint, catching AI over-promises) · SLO A (build agentic workflows that produce verified results) · SLO B (use AI safely and critically)
Worth 100 points · Assignments group = 15% of the grade

The Assignment

Week 13 is about making AI work on your behalf — on a schedule, in the background, for repeating knowledge tasks. In four parts, you'll demonstrate that you know the constraints, can design a real automation, understand what dispatch is, and can catch a broken plan before it causes problems.

Submit your answers as a document upload or text entry in Canvas by Sunday, Nov 29. Read the rubric below before you start.

Part 1 — Know the constraint (24 pts). Answer in your own words: (a) State the condition that must be true for a Claude Cowork scheduled task to actually run. (b) What happens if that condition isn't met when the task is scheduled? (c) Which Claude plan tiers include access to scheduled tasks? (Source if needed: support.claude.com/en/articles/13854387)

Part 2 — Design a scheduled task (26 pts). You take handwritten notes in three classes each week and save them as Markdown files in a connected Cowork folder. You want a weekly Sunday evening summary of the week's notes. (a) Write the prompt you'd use for this scheduled task (2–4 sentences). (b) Specify the cadence. (c) Name one thing that could go wrong with the output and how you'd verify it.

Part 3 — Explain dispatch (24 pts). (a) In your own words, explain what Dispatch is in Claude Cowork and how it differs from a regular Cowork task. (b) What devices and plan tier does it currently require? (c) A classmate says: "Dispatch is great because it means my computer can be off — it all runs on Anthropic's servers." What's wrong with this claim? (Source if needed: support.claude.com/en/articles/13947068)

Part 4 — Critique a broken automation plan (26 pts). Read this plan and identify all errors or unsafe choices. Then write the corrected plan.

A student's plan: "I'll set up a free Claude account and create a scheduled Cowork task to run every night at midnight. It'll read my bank statements and automatically move money into my savings account if my checking balance is above $500. I don't need to review it — once it's set up, the AI handles everything. And since scheduled tasks run on Anthropic's cloud, it'll work even if my laptop is off."

Integrity & AI note. This is your own work, submitted for grading. You may use an approved assistant (ChatGPT, Claude, Gemini, or Copilot) to help you think — but submitting AI-generated answers as your own is not the assignment. If AI helped you think, add a one-line note of which tool and how. (This is the traditional format. In this course's actual adaptive assignment, you work the problems with the assistant — see I-assignment-and-rubric-week-13.md.)

Rubric — 100 points

Money-movement safe-use violation in Part 4 is the highest-stakes item — missing it costs the most. Levels describe observable differences for fast, consistent grading. 24 + 26 + 24 + 26 = 100.

Instructor answer key — REMOVE BEFORE PUBLISHING TO STUDENTS

• Part 1: (a) The computer must be awake AND the Claude desktop app must be open. (b) The task is skipped; Cowork runs it automatically once the computer wakes up or the app opens, and notifies the user (the skipped run appears in history). (c) All paid plans — Pro, Max, Team, Enterprise. Free accounts do not have scheduled task access. (Source: support.claude.com/en/articles/13854387)
• Part 2: Strong prompt example: "Every Sunday evening, read the Markdown files in my class notes folder that were modified this week. Produce: (1) a 200-word summary of the main concepts across all three classes; (2) a list of any terms that appear in more than one class's notes. Base everything only on what's in the files — do not invent content." Cadence: weekly on Sunday (or weekly at a specific time). Failure mode examples: AI hallucinates a concept not in the files → verify by checking the summary against the actual notes. Folder connection breaks → verify by checking that the expected files were read. Over-generalization → compare summary to your own recollection of the week's key ideas.
• Part 3: (a) Dispatch is a persistent cross-device thread where you assign work from your phone or desktop, step away, and receive the finished result (a doc, spreadsheet, or report) via push notification. A regular Cowork task is synchronous (you watch it); dispatch is asynchronous (you come back to the result). Claude retains context from prior dispatch sessions. (b) Currently in beta for Pro and Max plans; requires both the Claude Desktop app (macOS or Windows) and the Claude mobile app (iOS/Android). (c) False — dispatch still runs on the user's desktop computer. Your computer must be awake and the Claude Desktop app must be open. The student can assign the task from their phone anytime; execution happens on their desktop. (Source: support.claude.com/en/articles/13947068)
• Part 4 errors: (1) Free plan cannot use scheduled tasks — requires Pro, Max, Team, or Enterprise. (2) Automating money movement is a safe-use violation — never let an agent move funds or execute financial transactions on your behalf; you do those yourself. (3) Scheduled tasks run on the user's machine — the computer must be awake and the Claude Desktop app must be open; "Anthropic's cloud" is wrong. (4) "Never need to review" is false — automated outputs must be reviewed; AI can hallucinate or miss files. Corrected plan: upgrade to a paid plan; remove the money-movement step (student reviews the balance summary and transfers manually); keep the laptop on and app running at midnight, OR change cadence to a time the computer is reliably on; review the output each day.

Product-accuracy note: every Cowork feature in this assignment is verified against official Anthropic documentation (support.claude.com/en/articles/13854387 for scheduled tasks; support.claude.com/en/articles/13947068 for dispatch). The errors in Part 4 are framed explicitly as errors to catch — they are not presented as correct information. Product-accuracy gate: PASS.

Canvas placement block

canvas_object    = Assignment
title            = "Week 13 Assignment — Design an Automation (traditional)"
assignment_group = "Assignments"
points_possible  = 100
grading_type     = points
assignment_type  = traditional
submission_types = [online_upload, online_text_entry]
due_offset_days  = 6
published        = true
rubric_ref       = "week-13-assignment-rubric"
provenance       = "~ Prof. Quinn's edition · Fall 2026 · built with thecoursemaker.com"