Month four of the three-day role pilot at Hantam Service Cooperative, a fictional shared-services cooperative in the Northern Cape, was the hardest. Month-end landed in a week with a public holiday. A member disputed a large allocation on the same day. The person who normally backed up the anomaly rota was on leave. Nobody missed a deadline, and on the surface the pilot looked fine.
The monthly review told a different story. Naledi, the analyst on the pilot, had logged five hours on a Thursday she was not contracted to work. The platform engineer on the anomaly rota had absorbed about five extra hours a week for a month. Two handover notes were missing. None of these was a crisis. All of them were early signs of the ways shorter work fails.
This final part of the series is about those failures. Reduced-time roles do not usually collapse. They erode, quietly, until the role is a full-time job with less pay or the colleagues around it are carrying the difference. The organisations that make them work are the ones that look for the erosion early and are willing to end a pilot honestly if it cannot be fixed.
Six ways shorter work fails
The failures are predictable, and each has a recognisable early signal.
1. The gain was never real. The time saving was estimated, not measured, or measured in a good month. People are poor judges of their own speed with AI tools, a point covered with evidence in Faster emails are not a business case. If the role was cut to fit a gain that does not exist, every other failure follows.
2. Coverage gaps. Work that needs someone every day was left with a person who is there three days. The signal is items waiting for the role-holder's return that should not have waited, and cost anomalies acknowledged late.
3. Meeting gravity. Decisions drift back into synchronous meetings that reduced-time staff cannot attend. The signal is decisions the analyst learns about after they were made, and decision records written after the fact or not at all.
4. Work intensification. The role-holder absorbs the gap personally: longer working days, work on non-working days, skipped breaks. The signal is logged hours above contract and activity on non-working days. The UK four-day-week pilot report found signs of it: perceived changes were roughly evenly split, with 36% of employees reporting higher work intensity and 31% lower and a small, weakly significant rise on average, and interviews in one larger company that had prepared little found staff describing their work as a battle against long to-do lists (full report, February 2023).
5. Inequity. The benefit goes to some staff and the cost to others: colleagues who receive moved work without capacity, or shift-based teams who cannot share the arrangement and cover for those who can. The same UK report recorded senior managers at two companies saying a permanent move might involve more reliance on subcontractors who would not themselves work a four-day week. That is the inequity risk in plain sight: shorter work for some, supported by others who keep a five-day week.
6. Coordination debt. The handover notes, decision records and queue described in Part 4 decay under pressure. The signal is missing or thin handover notes and failed Monday-test drills.
Signals, measures and first responses
A pilot needs to watch for all six from the first week, with a first response agreed in advance so that nobody has to improvise under pressure.
| Failure | Early signal | Measure | First response |
|---|---|---|---|
| Gain not real | Kept work does not fit the hours | Logged hours on kept work against plan | Re-measure across a full cycle; reopen the scope ledger |
| Coverage gap | Items waiting for the role-holder | Queue items past due on non-working days | Move the duty to the rota; name a backup for the backup |
| Meeting gravity | Decisions made without the analyst | Decisions without a written record | Move or replace the meeting; enforce written decisions |
| Intensification | Hours above contract; non-working-day activity | Weekly logged hours; after-hours activity | Conversation within a week; remove scope, do not ask for effort |
| Inequity | Colleague load rising | Hours and queue for receivers of moved work | Rebalance or fund the moved work explicitly |
| Coordination debt | Missing handovers; failed drill | Handover completion; Monday-test pass rate | Reset the template; protect time to write it |
Two rules govern the responses. First, the response to a signal is a design change, never a request for more effort from the person. If the role-holder is working extra hours, the role is too big. Second, signals are reviewed by someone with the authority to change the design, not only by the person experiencing them. At Hantam the monthly review is attended by the analyst, her manager and the finance director.
What Hantam did in month four
Each signal from the hard month led to a specific change.
- Naledi's unrecorded Thursday was treated as a design fault. The review found the cause: the disputed allocation needed her judgement and the dispute process had no rule for non-working days. The fix was a dispute rule: acknowledgement by the rota, investigation on her next working day, with the member told when to expect an answer. Her extra hours were paid.
- The platform engineer's extra load came from anomaly alerts that were noisy rather than serious. The team grouped related alerts and raised the threshold for low-value ones, which cut his rota time back to plan. His manager agreed the remaining rota time as a funded part of his role, not a favour.
- The missing backup led to a named second backup for the anomaly rota during leave periods.
- Public-holiday weeks got their own rule: in a week with a public holiday on a working day, handover notes move a day earlier and the overlap meeting moves with them.
- The two missing handover notes were written up in the review and the AI-drafted template was shortened, because the analysts said the long version was what got skipped on busy days.
Classifying each problem before arguing about it, in the way Classify the incident before you argue about it recommends for technical incidents, kept the review from turning into a debate about whether the pilot was a good idea.
Exit criteria, agreed at the start
A pilot without exit criteria never ends. It drifts into permanence or is quietly abandoned. Hantam agreed four criteria before the pilot began, and wrote them into the decision record from Part 1.
- Outcomes held. Allocation accuracy, report timeliness and member query response times at or better than the full-time baseline.
- The role-holder's load is sustainable. Logged hours within contract in at least five months of six, and no recurring pattern of non-working-day activity.
- Colleague load is sustainable. People who received moved work are within their own planned hours.
- The role-holder wants to continue. Asked privately, not in a meeting.
If all four hold, the arrangement continues and is reviewed annually. If one or two fail and the cause is fixable, the pilot is adjusted and extended once, by agreement. If the outcomes fail, or the load cannot be made sustainable, the pilot ends and the analyst returns to full-time work under the right agreed at the start, or chooses another arrangement she prefers.
Ending a pilot is a change to someone's working arrangements. It should follow what was agreed in writing at the start, and any change beyond that needs the employee's agreement. HR and employment-law advisers should be involved in both the original agreement and the exit. This is general information, not legal advice.
Hantam's pilot met all four criteria at month six, after the month-four changes. It continues. The cooperative also recorded what it would have done if it had not, because a decision that could only ever go one way is not a decision.
Failure does not settle the question either way
A failed pilot does not prove reduced-time roles cannot work, any more than a successful one proves they always will. Most failures trace to a specific design gap: a duty left uncovered, a meeting culture left unchanged, a gain that was assumed rather than measured. Fix the gap, and the same role may work.
Equally, success in one role at one organisation does not make a programme. The UK pilot's own findings were reported by organisations that chose to take part, and the three companies that paused the four-day week are as informative as the 56 that continued. Each new role needs its own baseline, its own scope ledger and its own exit criteria. The benefits ledger is a sensible place to record each decision, and What an AI value dashboard should show the board describes how to report the results upwards without turning them into a single success story.
That is the thread through the whole series. When AI reduces the effort a set of outcomes needs, someone decides where the gain goes. Shorter work is one legitimate destination among several. It is not automatic, it is not free, and it works only when it is designed, measured and reviewed as carefully as any other change to how an organisation operates.
Key points
- Shorter work usually erodes rather than collapses. Look for the erosion from the first week.
- Six failure modes are predictable: a gain that was never real, coverage gaps, meeting gravity, intensification, inequity and coordination debt.
- Respond to every signal with a design change, never a request for more effort.
- Agree exit criteria before the pilot starts: outcomes, the role-holder's load, colleagues' load and the role-holder's wishes.
- Neither failure nor success settles the question. Each role needs its own baseline, design and review.
This is the final part of The AI Productivity Dividend.
How CloudNala can help
CloudNala helps organisations run reduced-time and AI-enabled work pilots that can end honestly. We set up the six failure signals and their measures, agree exit criteria with leaders and the people in the pilot, run the monthly design review with you, and help you record and report the result, whichever way it goes. Where terms of employment change, we work alongside your HR and employment-law advisers.
Work with CloudNala
CloudNala helps organisations move from technology ambition to practical execution across cloud, AI, data, platform engineering and digital services.
Whether you are exploring AI, modernising your cloud environment, building a public-sector digital service, or turning an idea into a working MVP, we can help you shape the roadmap and deliver the next step.
Talk to us about AI and work design or write to us at consult@cloudnala.co.za