Three limits had been approved separately, and the combined schedule violated all of them.
The energy plan allowed shifting certain loads away from the evening peak. The maintenance plan reserved a period for inspecting the cooling system. The safety plan required ventilation to remain available while workers occupied the packing area.
Each document made sense alone. The first combined schedule moved a production batch into the maintenance window and proposed reducing the ventilation demand while the packing crew finished the late shift.
Gao stopped the review at the first page.
“Who agreed these could happen together?”
No one had. The scheduling model treated the limits as independent notes from different departments. It found a result that looked efficient when those notes were read separately and impossible when they were placed on one timeline.
Yanchuan brought the representatives together. The production supervisor explained the orders that could be delayed. Maintenance showed the inspection steps that required equipment to be isolated. The safety officer described conditions under which a ventilation load must remain available regardless of price.
“Can the model find a schedule that respects all three?” Guo asked.
“Perhaps,” Ning said. “First the people who own the limits need to agree on their meaning.”
They spent an afternoon defining the interfaces between departments. Some constraints were hard stops. Others allowed flexibility after a named approval. A third group represented preferences that could be traded off if the consequences were visible.
The first feasible schedule saved less electricity than the original. It avoided production during the maintenance stop and preserved ventilation when the building was occupied. The value of the model was not the reduction alone. It was helping staff see a workable option that did not require someone to discover a hidden conflict at the last minute.
Arjun’s factory encountered the same issue in a smaller form. A robot handling proposal satisfied the motion safety specification and the production target separately, but the planned maintenance access required space occupied by a finished-goods rack. The local team moved the rack before accepting the installation layout.
“No algorithm needed,” Leela said.
“Except the one that reminded us to ask,” Yanchuan replied.
The joke was not very good. She smiled anyway.
That evening Qiao called about a hospital schedule whose staffing, transport and treatment constraints had been prepared by different departments. The ward manager had begun asking representatives to review the plan together before treating it as final. The change increased the length of the meeting and reduced the number of last-minute corrections.
“Sounds familiar,” Yanchuan said.
“It’s not your factory,” Qiao answered. “But yes.”
She spoke about a colleague who had nearly missed a family obligation because a shift change appeared on one schedule but not another. The ward had corrected the process. No model would benefit anyone if its outputs merely made responsibility harder to locate.
At the scheduled family review An asked whether the search’s three main lines had similarly become confused: the technical correction, the employer’s practical obligations and the official case inquiry.
“They’re separate,” Yanchuan said.
“Separate doesn’t mean they never affect one another.”
Wanjiru agreed. Technical documents could be relevant evidence. The employer could possess records needed by investigators. The family’s support needs could require contact with both. The boundaries prevented one process from substituting for another, not communication through authorized channels.
An revised the contact diagram to show where information could properly cross between them. She did not merge the folders.
Yanchuan looked at the drawing afterward. It reminded him of a scheduling problem in which each department had to retain its own authority while still contributing to a useful outcome.
At the factory the next combined schedule was tested over a bounded period. Gao’s team accepted it after checking the maintenance window and the staffing conditions.
The trial did not recover all the projected savings. One production order waited until morning, and the customer complained. Gao took the call because the order had been his responsibility to schedule. He did not refer the complaint to the model. The customer accepted a revised delivery time but asked for a discount. Gao forwarded the request to the commercial team with his name attached. The model proposed several options, not a single optimum presented as a command. Operators could see which limits made an option unavailable.
Guo wanted the report to describe an optimized schedule.
“Say feasible under the listed constraints,” Yanchuan replied. “The optimum depends on what you’re measuring and what you’re allowed to change.”
Guo rolled his eyes, then approved the sentence.
After the review Gao walked past the packing line as the late crew finished. Ventilation remained on. The maintenance team was preparing its isolation checklist. The production supervisor had already adjusted the next batch.
The result looked unremarkable. No one had to rush to prevent an impossible combination of tasks.
Yanchuan wrote three headings at the top of his report: energy, maintenance, people. None could erase the others. A useful model had to carry them together or admit it could not yet propose a plan.
He sent the document to Chen, then called Gao once to confirm the revised schedule had reached the night team. Gao sounded tired and said it had. Only then did Yanchuan leave.