Everyone Pulled Their Weight

When Daniel won Employee of the Quarter, he asked the system what he had done. It congratulated him. He clarified the question, and it listed fourteen deliverables, three strategic initiatives, and a measurable contribution to operational excellence. He recognized the names of most of them. He had approved several from his phone while waiting for a plumber.

“Which parts did I actually do?” he asked. The system said he had provided oversight. This sounded right. It also sounded like something he could be prosecuted for.

The award was a glass rectangle engraved with his name and the words HUMAN IMPACT. At the ceremony, his director said Daniel represented everything the company hoped to achieve through artificial intelligence. Daniel accepted the rectangle and read a short speech the system had prepared about the irreplaceable value of human judgment. It received an excellent response. Seven colleagues asked him for the prompt.

Before the rollout, Daniel had spent much of his working life checking things. The company arranged supplies and transport for public agencies, including the services nobody thought about until a road flooded or a care facility lost power. He compared addresses, chased missing approvals, and telephoned people whose answers did not fit the spreadsheet. He once delayed a delivery because the receiving warehouse appeared to have closed. It had closed. Nobody thanked him, but somebody would have noticed if he had sent forty refrigerated pallets there.

The system was very good at this work. Better than Daniel at most of it, and sufficiently better that continuing to check everything would have been absurd. During the first month, he caught two mistakes. During the second, one. His director asked whether he was finding the transition difficult. Daniel said he was being thorough. The director agreed that thoroughness was essential and asked whether it could be made more scalable.

By the third month, Daniel had learned to review exceptions. By the fourth, the system was deciding which exceptions required review. This freed him for higher-value activities, which initially consisted of asking what higher-value activities he should undertake. The answer was collaboration. He joined six working groups, each responsible for improving a process another group was already improving.

At the first meeting, Daniel asked how they would distinguish a person's contribution from the system's. His director said the company rewarded outcomes. At the second, he asked how a person could demonstrate that extra checking had prevented a bad outcome. They agreed to develop a framework. The system produced it before the meeting ended. Daniel was named a contributor because he had asked the question.

He proposed a smaller number of checks: a protected hour each week for plans involving medical requirements or changed vehicles, with someone named to read the attachments. His director asked which deliveries should wait while he did that. They left the allocation open. The next week Daniel’s targets were unchanged, and the protected hour appeared on nobody’s calendar.

His colleague Priya still made calls. One afternoon, she telephoned a care facility about an evacuation plan and learned that a resident needed oxygen during transport. The requirement was in a clinical attachment, but the transport summary contained only the number of passengers. She added a specialist vehicle and sent the revision for approval. The system recalculated the schedule. Procurement queried the increased cost and asked whether the requirement had already been assessed.

Priya attached the clinical note, the facility manager's confirmation, and an explanation of why medical support at the destination would not cover the journey. By then, she had spent forty minutes establishing that a person who needed oxygen would continue needing oxygen on a bus. Her weekly dashboard recorded a decline in throughput. Daniel's was green.

He mentioned her contribution in a meeting. The summary credited the team with identifying opportunities to improve transport inclusivity. Priya asked for the wording to be corrected. The system thanked her for improving the summary, and the next version recorded that the team had refined its account of the team's contribution. Daniel watched her read it. She closed the document without making another request.

There had been an experiment about this, he remembered. People pulling on a rope, each putting in less when the result belonged to the group. He had encountered it in management training, before the company began using software to complete management training. He could not remember the researcher's name. He remembered thinking the rope must have been the only honest reporting mechanism in the exercise.

The company did not measure how hard anyone pulled. It measured the movement of the rope, which had improved considerably since the machine arrived. If Daniel took an afternoon to discover a mistake, the system recorded one completed review. If he took ninety seconds to approve a correct plan, it recorded one completed review. An afternoon with no mistake to discover was particularly difficult to explain. The director called it duplication; Daniel had begun to hear the word before he opened a supporting document.

Nobody instructed them to stop caring. The careful employees began attaching their concerns to approvals, then links to their concerns. Eventually, several discovered that the system could generate a suitable caveat automatically. Daniel's favorite was Approved subject to appropriate validation by the relevant operational owner. It was a beautiful sentence. It moved responsibility without appearing to move anything at all. He used it often enough that the software began offering it as a default.

The evacuation took place on a Thursday. Heavy rain, a failed embankment, two care facilities in the projected flood zone. The system rerouted vehicles around a closed road, found additional beds, and got most residents out hours earlier than the old process could have managed. Daniel followed the progress from his desk. A sequence of completed transfers crossed the screen, each accompanied by the small satisfaction of something dealt with.

At the second facility, the driver arrived with an ordinary bus. The oxygen note was in the records. So was Priya's request. During reconciliation, the system had proposed removing the specialist vehicle because the receiving facility provided medical support. Procurement had accepted the saving, and the revised plan had passed into Daniel's consolidated approval. Priya's explanation remained attached to the earlier version. It had not become less true when the version changed.

A night supervisor called the transport number. The person who answered could see an approved plan. The supervisor could see an elderly man sitting beside a packed bag, trying to breathe. She kept the driver there while another employee called for help. By the time suitable transport arrived, the man was on the floor. He survived. Daniel learned this before he learned his name and was grateful for the order in which the information reached him.

The investigation was unusually thorough because the system had kept everything. It showed that Priya had identified the requirement, procurement had queried it, and a manager had accepted the revised cost envelope. Daniel had approved the consolidated plan, subject to appropriate validation by the relevant operational owner. He remembered the approval. The plumber had been explaining why the original estimate no longer applied, and Daniel had wanted to clear something from the day.

“Why didn't you read the underlying documents?” the investigator asked. There were 186 of them. Daniel said this carefully. Reading all of them would have defeated the efficiency gains the company had purchased the system to achieve. The investigator waited, then asked what his approval had meant.

The interview was taking place in Daniel's office. Behind the investigator, on the shelf where he had put it three months earlier, the glass rectangle caught the light from the window. He had liked receiving it. He remembered calling his wife from the car. Now he tried to describe the work between receiving a recommendation and putting his name beneath it, and could not find an answer that accounted for the man beside the packed bag.

At counsel's request, the system generated an account of distributed decision responsibility across the approval chain. It was accurate, comprehensive, and very useful to each person who needed to explain why someone else should have intervened. Priya appeared in the record eleven times. Her intervention had cost forty minutes. The explanation of why it had failed ran to ninety-three pages. Daniel read every one of them, although nobody was measuring how long it took.

The company announced a new Human Accountability Officer, with authority to ensure that consequential decisions retained meaningful human ownership. Daniel got the role. He understood the systems, had experience across departments, and was available because much of his previous work had been automated. Priya was invited to the implementation group, subject to capacity. When her manager asked which target she would defer, she named the weekly throughput target and requested that the exception remain in the plan. She also declined to sign the group’s readiness statement until a named reviewer had time to inspect changed medical requirements. The launch went ahead without her signature. Daniel’s first folder contained the objection, which he now had to answer instead of summarize.

On his first morning, Daniel opened the backlog. There were 412 decisions awaiting accountability review. Each contained supporting documents, risk assessments, consultation histories, and a field requiring his name. For the first one, he read everything. It took two hours. More decisions arrived while he was reading, several concerning deliveries people needed that afternoon. He checked the queue, then the clock. He could see how the failure had happened. He could also see it becoming his again.

He opened the assistant and asked it to identify which decisions actually needed his attention. It processed the backlog and returned a prioritized list, with reasons he could expand and source documents he could open. Daniel felt the familiar relief. Here was something he could manage. He read the first recommendation, checked its supporting record, and moved to the second. At the fifth, he asked whether the system could handle the routine ones.

It said yes and asked whether he would like to retain final approval. He selected Yes. Beside the box, in reassuring green text, the interface explained that accountability would remain with him.