The robot’s arm stopped two centimetres short of the tray.
The demonstration engineer pressed a button, checked the safety enclosure and restarted the sequence. On the second attempt the arm reached the correct position but gripped the wrong corner of a component. The tray shifted, and the operator stopped the motion before anything was damaged.
The supplier looked embarrassed. Arjun looked interested.
“What happened?” he asked.
“The part position varied,” the engineer said. “We can tune the vision system.”
“Who will tune it here after you leave?”
The room went quiet. The demonstration had been organized around throughput, precision and labor saved. Nobody had prepared a maintenance plan for the factory’s actual mix of parts and tools.
Ning examined the proposed cell. It required a space cleared of other equipment and a dependable supply of compressed air. Both were available on the drawing. On the floor, the air line served several existing machines, and the proposed safety perimeter would obstruct a route used during shift changes.
“You don’t need a better robot to fix this first,” she said.
The supplier offered a more expensive model with additional sensing. Leela asked how quickly a damaged gripper could be replaced and whether the factory’s technicians would be authorized to service it without voiding a warranty.
The answer depended on a training package and parts contract that had not been included in the quote.
Yanchuan made three columns: robot purchase, workshop adaptation, continuing service. The first was the smallest part of the uncertainty, though it had occupied almost all of the brochure.
Arjun told them that the task under consideration involved moving warm metal parts between two approved stations. Repetition was tiring. Workers had asked whether the job could be made safer, not whether the factory could replace them.
“Then measure the risk and the work first,” Ning said. “We might use a robot, a mechanical aid or a change in the line.”
They observed the shift with permission. The parts arrived at varying temperatures, and the operator sometimes waited for a batch to cool before handling it. The production schedule treated each cycle as identical. It was not.
Leela explained that operators recognized certain unusual sounds and stopped the line before a part was handled. A robot could be given sensors and stop conditions, but those requirements had to be specified, tested and maintained. No one should assume a machine automatically inherited the judgment of the person whose motion it copied.
Back in Shenzhen Guo wanted a preliminary estimate of how quickly automation would pay for itself. The first calculation looked attractive when it counted the operator’s handling time as fully replaceable labor. It looked different when it counted the operator’s other tasks, the continued need for supervision, the new maintenance role and the space displaced by the cell.
“Does that mean we shouldn’t buy it?” Guo asked.
“It means we haven’t chosen the solution yet,” Yanchuan replied.
Mei suggested testing a simple mechanical handling aid as a comparison. It might reduce strain without the software and specialized parts of the robot cell. The local factory team agreed to try it under their safety procedure.
The aid was less impressive on video. It did not move by itself. The worker guided the part with reduced effort, while maintaining the ability to inspect it and stop. A maintenance technician could repair most of the mechanism using tools already on site.
The alternative did not prove robots were useless. It showed that a particular problem could have more than one responsible answer.
Arjun asked the supplier for a revised proposal with the required workshop changes and service commitments included. He also asked the staff whether they wanted the trial to continue. Several did. Others preferred developing the mechanical aid. They agreed to compare the options by safety, workload, maintainability and total cost over time.
On a call An asked Yanchuan why he sounded pleased that an expensive robot might not be purchased.
“Because the problem could still be solved,” he said.
“Then your customers are paying for the right problem.”
He thought of his brother insisting on examining mislabeled crates. The object in front of him was not the entire system. It never had been.
A month later the supplier returned with a more realistic proposal. It included training for local technicians, a spare-parts schedule, approved maintenance access and a modest target for throughput. The factory kept both options open.
The robot waited in its demonstration enclosure. The mechanical aid had already completed an approved trial.
Yanchuan closed the comparison without declaring a winner. A machine could be clever, fast and expensive. It could not operate as a useful service without a place that could keep it working.
Arjun asked Leela to list which repairs the local team could perform without waiting for the supplier. The answer would decide whether the robot returned for another trial.