ROI & the time audit
The honest math: what a task really costs, what an automation really costs (build AND upkeep), and the one-week time audit that finds money lying on the floor.
Automation has real ROI and inflated folklore. The honest math keeps you from both errors — automating things that don't pay, and not automating things that obviously do.
What the task costs (it's more than the minutes)
- The visible time: minutes per run × runs per week. Riley's ticket copying: 4 min × 50/week ≈ 3.3 hrs/week.
- The switching tax: every interruption costs recovery time on whatever you were doing. Frequent small tasks are worse than their minutes — this is why the drip beats the puddle even at equal totals.
- The latency cost: work waiting for a human happens later. The urgent complaint sits 3 hours because Riley was in a meeting. Automation's speed often matters more than its labor savings — a 2am ticket gets logged at 2am.
- The error rate: manual repetition has a defect rate (typos, skipped rows, wrong tabs). Honest accounting counts cleanup time too.
What the automation costs (it's more than the build)
- Build: for the workflows in this course, hours — not weeks. Your first will take an evening; your fifth will take a lunch break.
- Subscription/usage: platform plans and (for AI steps) per-run model costs — a 'model' being the AI engine that powers each AI step, rented per use — typically cents per run; know the number anyway.
- Maintenance: the underestimated one. Connected apps change, fields get renamed, edge cases appear. Budget roughly an hour a month per running workflow for care and feeding. Ten workflows ≈ a real chunk of someone's job — which is why Module 5 makes every automation carry a named owner.
The rule of thumb that survives contact with reality: an automation should repay its build time within a month, or be strategically important enough that you'd defend the exception out loud. Riley's 3.3 hrs/week chore against an evening's build repays in under a week. That's the shape of a first project.
For one week, keep a note open. Every time you do something repetitive, log four words and two numbers: what, for whom, minutes, and a gut 1–5 on 'could a recipe do this?' No analysis during the week — just capture. People predict their top automation candidates before auditing and are wrong about half the time; the audit finds the invisible drip (status update someone asks for daily, the re-formatting nobody counts). This is Lab 1's raw material — start today.