Month 4 · Design automation before acting

Complete fictional source F4

P01 is an invented stationery item with reference stock 20 units. Each row below is a separate test, not a shared inventory transaction. There are no prices, customers, delivery commitments or connected business tools.

Input order is significant: the first E01 establishes the comparison for the two later E01 rows. Every result is simulated; approval, action execution and stock changes stay false.

Week 1 · Trigger, condition and action

Goal: Define what starts a rule and what it may actually do.

A trigger is an observable event, not permission for every action. Each fictional request contains an event ID, request ID, product reference and quantity. A condition selects a branch. The proposed action is a draft awaiting human review. Drafting is different from sending a reply, reserving stock or confirming an order. Write these parts separately so another person can inspect the scope.

Source F4 contains one product P01 with reference stock of 20 units. Every row is an independent test against that same source. Do not combine quantities from two tests and claim both can be fulfilled: this simulator never reserves or subtracts stock. A working service would require authorized concurrency controls and a current source, neither implemented here.

An identifier recognizes a repeated event without a person’s name or phone number. Keep only the minimum fictional information. Quantity must be a positive integer and the product must exist. Missing information must stay missing. When a check fails, describe the unresolved point and stop before any external action.

Case for review

A proposed rule says: “Whenever a request arrives, automatically confirm it is ready.” What evidence and permission are missing?

Your activity

Rewrite the rule as trigger, ordered conditions and a draft pending review. Name three prohibited actions. Classify all nine F4 rows before reading the answer.

Download worksheet 1 (CSV)

Supporting artifact 1: scope contract and initial classification.

Week 2 · Order the conditions

Goal: Build a decision that another person can explain and repeat.

Order matters. Identify the event, then check repetition or conflict, then source availability and content. A small quantity never implies approval. An incomplete request, an unknown product and insufficient reference capacity require different corrections. None independently proves that a person did something wrong.

For P01, quantity 20 meets the boundary; 30 exceeds it by 10. This only selects test branches, not commercial availability. Compare quantity less than or equal to reference stock in the same units. Do not compare number strings as text or accept zero as a valid requested quantity.

Human approval is a separate decision that would need an authorized person and evidence in a real system. This exercise has no approval button and no branch changes approval status. A draft may list checks still needed. It must never invent an approver’s name or claim a manager already approved. The workflow ends before external interaction.

Case for review

E02 requests 30 units and E08 requests 20. A draft claims both fit “because they are the same product”.

Your activity

Draw the branches. Compare 30 with 20 and 20 with 20 explicitly. Define output for zero quantity, empty reference, unknown reference and unavailable source.

Download worksheet 2 (CSV)

Supporting artifact 2: decision diagram and boundary checks.

Week 3 · Duplicates, conflicts and failures

Goal: Avoid repeated actions while preserving incompatible changes.

An exact duplicate has the same identifier and content. The fifth F4 row repeats E01 with quantity 5 and is ignored. The sixth reuses E01 but changes quantity to 7: this is a conflict, not an authorized update. Silently accepting it as new could duplicate actions in a real service. Stop and preserve the distinction for review.

The simulator’s ledger lasts only for one local run. Running again starts a fresh test. This supports learning and reproducibility but offers no durable protection across computers, tabs or restarts. A production service would need authorized persistent storage and atomic controls. Do not describe this example as an operational integration.

An unavailable source does not mean zero stock. E09 requests 5 but sourceAvailable is false. The correct statement is “unable to verify”. Blind retries may repeat actions. Describe what evidence an authorized person would need and retain only relevant error information. Never record passwords or real conversations in this exercise.

Case for review

A proposed workflow labels E09 “sold out” and treats the sixth E01 row as a new order.

Your activity

Correct both mistakes. Build a nine-row matrix of input, expected outcome, observed outcome and explanation. Preserve the original row when distinguishing a duplicate from a conflict.

Download worksheet 3 (CSV)

Supporting artifact 3: nine-case matrix and two documented corrections.

Week 4 · Run the tests and assemble M04

Goal: Submit reproducible evidence without confusing a simulation with operations.

Use the laboratory after writing expected outcomes. Compare all nine results and explain discrepancies. No network or progress storage is used; reloading removes the output. Keep a local worksheet if you want evidence. A screenshot can show a result, but the inputs and explanation are needed for reproducibility.

Combine the four supporting artifacts into one monthly activity M04: F4 source, scope contract, decision diagram, test matrix and corrections. Label the version and unresolved questions. Weeks are relative self-study blocks. No hours or completion date are guaranteed. This monthly package does not replace the other eleven activities or final portfolio.

Use approved policy v1 for self-review: facts and sources; artifact and process; data and human review; reasoning and limits. For each criterion, 0 means absent or incorrect evidence, 1 partial evidence with unresolved points, and 2 sufficient justified evidence. The 6/8 minimum with no zero is a criterion for authorized human review, not a certification generated here.

Case for review

One self-review gives 2, 2, 2, 1 for a total of 7. Another gives 2, 2, 2, 0 for a total of 6.

Your activity

Run the tests and document why no real action occurred. Score the four criteria with references to artifacts. Describe a concrete correction for every partial criterion.

Download worksheet 4 (CSV)

Single M04 submission: bounded automation design, nine tests and justified self-review.

Assemble your single monthly package

Download M04 package guide (TXT)

Keep your work locally. Formal review requires actual artifacts and an authorized human reviewer. No certificate, enrollment or benefit is issued by this page. The other eleven months and final portfolio remain separate parts of the annual program.

Optional AI tutor — read consent and limits first · My personal study marks, not academic verification