Integrators are not usually the ones telling you to slow down. But the fastest way to waste an automation budget is to automate a process that was not ready for it, and we would rather say so at enquiry stage than at commissioning.
The process is not stable yet
Automation amplifies whatever it is given. If your manual process produces acceptable parts only because skilled operators are quietly correcting for variation — in the incoming material, in the fixture, in the previous operation — a robot will faithfully reproduce the variation instead of absorbing it.
Fix the process capability first. A stable manual process is a candidate for automation. An unstable one is a candidate for root-cause work.
The product is about to change
Hard automation designed around a part geometry that changes in eighteen months is a write-off. If a revision is coming, either wait for it or specify flexibility explicitly — quick-change tooling, parameterised programmes, extra reach — and accept the cost that flexibility carries.
The volume does not support it
There is a real floor below which automation cannot pay back, and it moves depending on part value, labour cost and quality risk. Low volume does not rule out automation, but it changes the answer: a collaborative cell that an operator shares may be right where a fully guarded line is not.
Nobody owns it internally
This is the one that kills more projects than any technical factor. A cell needs someone in the plant who understands it, is trained on it, and has the authority to keep it running. Where that person does not exist, availability falls quietly for months until the cell is bypassed altogether.
The best automation projects start with a plant that already knows exactly why its current process is limiting it.
What good readiness looks like
- A stable, capable manual process with known cycle time and scrap rate
- A product design that is frozen, or a clear specification of the variants to cover
- Volume and part value that make the payback credible without heroic assumptions
- A named internal owner with training budget and authority
- Honest data — real cycle times and real reject rates, not nameplate figures
Meet those five and an automation project becomes a straightforward engineering exercise. Miss two or more and no integrator, however good, will save it.
