Judge the two plans
Judge the two plans
Two colleagues propose ways to produce a weekly competitive digest: one email each Monday covering pricing, product and messaging changes across your market. Read both plans and judge which one will still work in month three.
Compare the plans against the workflow criteria, then choose the one you would put your name on.
Output A
Plan A — “Each Monday I will paste everything I have into one long prompt: the competitor pages, the newsletters, the pricing sheet, and ask for a digest of anything important. If the result feels thin, I will add a sentence like 'be thorough and cover all the changes'. I will send it if it reads well.”
Output B
Plan B — “The digest runs in four steps. Inputs are fixed: three competitor pages, their changelogs and the pricing sheet, saved in one folder each Friday. Step 1 extracts any change as a table row: source, date, what changed, quote. Step 2 filters to pricing, product and messaging — the three areas our team acts on. Step 3 drafts two-sentence entries from the filtered table. Step 4 is my review before anything sends: I check each entry against its quote. If a source is missing or unreadable, the row is flagged and the digest ships with a 'not checked this week' note.”
Rubric — answer each question for both outputs before deciding
- The inputs are named and arrive in a predictable state
- Each step does one job, and its output is the next step's input
- There is a specific check before the digest leaves
- A missed or broken input has a defined path instead of silent silence
Hint
Look for the check step — a place where a human or a rule compares the output to the source.
A plan that cannot fail visibly will fail invisibly: ask what happens on a broken week.
Why this is the answer
Both plans describe the same job; only one describes a process. Plan B makes each stage inspectable — extraction, filter, draft, review — so a mistake is locatable and fixable, and a missing source is declared rather than hidden. Plan A removes every seam, which feels efficient until the first wrong number ships. The workflow criteria are not bureaucracy: inputs, stages, checks and edges are exactly the places quality is either designed in or lost.