What to automate first · Scoring

Which of these six tasks first

Score six of the agency's candidate tasks on four multiplied factors, pick the first two to automate in order, and name the one that waits.

casemediumExercise · 3 min · 9 of 75

On this page

One builder, one quarter, two automations at most. The order is the decision, and a wrong first pick costs the quarter.

The situation

The agency from the scoring frame has 14 people: you as founder, staff at an assumed $30 an hour loaded. The lesson scored six of its twelve candidate tasks; these are the other six (the 92 leads a week are its 400 a month), and this quarter you will build two of them at most. Each source system a task reads from is one connector to build.

Task	/week	min	error cost	input
Invoice from timesheet	40	6	$50 rework	export file
Reply to inbound lead	92	10	$1,500 deal	form + email
Weekly client report	12	45	$200 credibility	6 dashboards
Apology to an unhappy client	1	30	$8,000 churn	free text
Screen CVs	30	8	$300 missed hire	PDFs
Categorise bank transactions	150	1	$20 mis-post	bank feed

Score each factor 1 to 5 on the lesson's bands: frequency in runs a week, cost in person-hours a week, tolerance from the cost of one error, input from how much of it arrives as fields.

The question

Pick the first two tasks to automate, in order, and one not to automate this quarter, with the reason for each.

A strong answer · Rubric · Where answers go wrong

Write your answer first, then open

A strong answer

Hours first: transactions 150 × 1 min = 2.5 h a week; invoices 40 × 6 = 4 h; report 12 × 45 = 9 h; CVs 30 × 8 = 4 h; leads 92 × 10 ≈ 15.3 h; apology 0.5 h.

Multiplied, frequency × cost × tolerance × input: bank transactions 5 · 3 · 5 · 5 = 375; invoices 4 · 4 · 4 · 5 = 320; report 3 · 5 · 4 · 5 = 300; CVs 4 · 4 · 3 · 3 = 144; leads 4 · 5 · 2 · 3 = 120; apology 1 · 1 · 1 · 1 = 1.

Transactions first, invoices second. Leads have the most hours and an hours-first list would start there, but a $1,500 error and half-free-text input put them fifth. The report is third, and six dashboards mean six connectors, the largest build on the list, so the build pass keeps it there. That pass may re-order the top three; it never promotes the bottom three.

The apology is not automated this quarter: one error costs $8,000 of churn and the input is free text. Either reason is enough alone; rarity only confirms them.

Done for the first pick: by Friday 17:00 every transaction in the week's bank feed carries a category, and those the model is unsure of, assume 5%, sit in a review queue under the operations manager's name.

What would change the order: if the bank feed needs a connector costing more than the invoice run's 12-hour build in the arithmetic of a task, invoices go first.

Rubric

scores all four factors for all six tasks and multiplies them, rather than summing or ranking by hours
shows the arithmetic for at least the top two, from weekly hours to the product
rejects the apology on its error cost and free-text input, not on frequency alone
applies build cost as a second pass that re-orders only within the top three
states "done" for the first pick as a condition someone else can check, with a time and an owner

Where answers go wrong

"Leads first: the most hours, and revenue." Tolerance is 2 and half the input is free text; one mis-sent reply costs a $1,500 deal. Its sums are good, which is why it comes back once the approval gates of module 4 catch a wrong reply before it is sent.

"The report before invoices: 9 hours a week against 4." This ranks by hours. The report needs six connectors before it saves a minute; the frame already puts it third and the build pass keeps it there.

"The apology waits because it is rare." The weakest of the three reasons. An answer resting on rarity would automate a frequent task with an $8,000 error and free-text input, which is exactly what the frame exists to stop.