Finance asks if automation pays by comparing “machine cost” to “operator wage.” The comparison is wrong from the start—not because payroll math is bad, but because a manual station’s real cost never fits on one salary line.
The hidden invoice of manual work
Repeat pick-and-place, load/unload, fastening, visual check—different job titles, similar cost structure:
Training and churn: three weeks to competence, quit before peak stabilizes. Hidden cost is the lead re-teaching, yield drift, and job ads
Consistency: the same motion differs Monday vs Friday, day shift vs night. Not always scrap—sometimes the 2% you cannot explain in a customer complaint
Speed ceiling: people can be urged, not duplicated. At peak, overtime hits limits—legal, physical, and “everyone leaves if we push again”
Space and takt: two-person stations wait, hand off, talk. Those seconds are not in standard time sheets, but they live in Cpk
Those costs sit in five departments, so no monthly meeting is shocked by one “manual bill.” Automation front-loads spend, so it looks loud.
When “flexibility” is an illusion
Manual’s great strength is fast changeover: new order, new method tomorrow. True—but only when changeover frequency is high enough to matter.
If the same motion repeats two thousand times a day for six months without a product change, “flexibility” means you pay a premium every day for a rare event—trading variance for occasional freedom.
A simple test: in the last twelve months, did this station stop more for changeover or for people variance? If the latter, flexibility is narrative, not data.
How automation should be compared
A fairer frame is not “arm vs wage” but two line hypotheses side by side:
Manual path: wages + overtime + training + average scrap + opportunity cost of turned-down peak orders Automated path: depreciation + one-time integration + power + maintenance + changeover prep + remaining labor (teach, patrol)
Many SMEs stop at integration because it is hard to estimate. Use a band: pessimistic, base, optimistic integration budgets and see if break-even still fits acceptable years.
Even a four-year pessimistic payback can be rational if hiring risk on the manual path is rising.
People are not deleted
We do not write automation as “replacing workers.” On most collaborative stations, people remain: exceptions, replenishment, first-article sign-off, process dialogue with the customer.
Automation takes the slice that is repetitive, provable, and unfriendly to joints. What remains is often more expensive, scarcer, and more experience-dependent—if the plant still hires “cheap repeat labor” with an old wage model, it fights its own automation investment.
To the plant manager still hesitating
If you stand at the line before lights-out and watch one person repeat the same pick-and-place, ask three questions:
Which seconds of this motion are wasted on the human body?
If orders double next year, where is week-one bottleneck—people or takt?
If this person does not come in tomorrow, how many days to restore capacity?
If two of three answers point to people, the automation conversation is mature—not because robots are sacred, but because you can see manual’s real price.
It is not one invoice. It is a chronic, scattered spend that only hurts in peak season. Seeing it takes more courage than reading an arm quote.
