Many pilots do not die in integration. Acceptance gets signed; cycles reproduce—but the production schedule often never names the arm. Ask the plant manager if it is “live” and the answer is “it runs.” Ask which option night shift defaults to and the room says manual—less hassle. What is missing is rarely torque or repeatability. It is an administrative seam: capex sits in digital transformation or equipment budget; accountability sits on production. One side needs a reportable milestone; the other needs a default capacity line planners dare to write. Both logics are valid. Together they leave the machine capitalized and the operator still on the job. The pilot lives between “good for demos” and “required for ship”—no one declares failure; no one is assigned to declare success.
“Buy one and try it”—without defining the try, two opposite errors both look reasonable
Without written acceptance bounds, “try” slides low and high at once. Too low: one good pick counts as pass; week one exposes missing seconds in takt or slow changeover; the project becomes “not yet.” Too high: month one must match manual yield; normal drift reads as immaturity; same shelf. Opposite verdicts, same landing: no graduation rule, no stop rule—only “keep observing” in meeting minutes.
More common and more wasteful: engineering passes, organization does not. EOAT, vision, takt are there; MES or the whiteboard schedule never marks the station robot-first. Integrator exit switches the support chain. With no internal owner, “run when needed” in plant language usually means manual forever.
The first ninety days after handoff are the line acceptance test
FAT, SAT, and training records answer delivery compliance, not shift habit. The real exam starts after the integrator leaves: pre-peak season, short night crew, emergency changeover at 2 a.m. Without one named station owner, friction returns in concrete forms—who signs a taught offset, whether a few points lost is tuning or manual bypass, whose P&L eats downtime while spare parts clear procurement. You bought hardware; you did not buy the operating mechanism. The invoice will be paid from output, slowly.
Showing automation to visitors is not the sin. Sin is when site selection mainly serves “visible from the door, stops without blocking the main line.” Over time the robot becomes tour equipment. If the pilot must scale, location should answer one question first: is this where repeat labor hurts most right now—not where it bothers the meeting mood least?
Graduation is not “no faults”
Graduation is earthy: without the integrator watching, planners still schedule this station. That usually takes half a page at charter—not another arm: takt, yield, changeover time, continuous run hours as numbers; named owner and teach rights; how manual bypass stays; within thirty days after handoff, expand or stop.
If you already own a cobot that demos but does not ship, pricing a second unit is rarely the urgent move. Check three things first: can you find the document that said it passed; who may reset at night; if this were the only unit forever, does this location still justify it?
