English (en)
Using AI all the time taught me to delegate to people too — What I learned from a ¥3,000 moving-packing job
Five-second answer: Daily AI use can build a habit of asking “Who should do this?” before automatically doing everything yourself. That became obvious when a ¥3,000 request for packing help quickly drew three applicants.
Thirty-second overview: Packing small items is easy but tedious. A local listing offered two to three hours of work, ¥3,000 or a meal of similar value, pickup for someone nearby, and air conditioning down to 18°C. If nobody replied, the work would simply continue alone. Several people did reply. Choosing among them showed that ratings are not enough: direct answers, useful suggestions, short travel distance, and low coordination effort matter too.
Before a move, the worst task was not lifting furniture. It was building boxes, putting small objects inside, and writing labels.
It was easy. It was repetitive. That makes it a perfect task to hand off.
1. A ¥3,000 listing quickly drew three people
One professional service currently lists small packing jobs from ¥11,000 and two workers for two to three hours from ¥22,000.[6]
The whole service was not needed. Only an extra pair of hands was needed.
The local listing said:
- about two to three hours;
- ¥3,000 or a meal worth about the same;
- pickup if the person lived nearby;
- no heavy furniture;
- air conditioning could go down to 18°C.
That last line was strangely specific, but it made the working conditions clear.
If nobody applied, packing would continue alone. Instead, one person wrote, “I’d love to do it.” Then a second person appeared. Then a third.
Someone even asked whether weekend time was available, so a morning slot was opened. The person posting the job had somehow become the manager of a one-person shift schedule.
The candidates looked like a game party: an ordinary guy, a muscular guy, and an older man who seemed to have walked out of a fantasy village. A packing request had turned into adventurer recruitment.
The test was easy to run because failure was cheap. Posting took a few minutes. No applicants meant no change. Money would be spent only after a useful match appeared.
Entrepreneurship research describes a similar approach: experiment, stay flexible, and risk only what you can comfortably lose.[3]
Instead of wondering for hours whether the idea could work, reality answered the question.
2. Every AI request is practice in giving clear work
Using AI repeatedly means practicing the same steps:
- Decide what “done” means.
- Break the job into smaller parts.
- Choose what the AI may do and what judgment stays human.
- Give limits and acceptance rules.
- Check the result and correct it when needed.
That is closer to delegation than to casual question asking.
A 2026 paper on AI delegation argues that complex work needs clear roles, boundaries, responsibility, trust, and verification—not just task splitting.[1]
Another 2026 study found that work-focused AI use was more strongly associated with giving instructions, constraints, and criteria before execution.[2]
This does not prove that heavy AI use directly causes better human delegation.
Still, repeating this process many times a day can reasonably build a new reflex: before taking a task yourself, ask who or what can handle it.
3. The new order is “AI, then a person, then me”
The operating rule is simple:
Can AI do it? If not, can a person do it? If not, I do it.
Writing, research, comparison, organization, coding, and drafting fit AI well.
AI has no arms, so it cannot place real objects into boxes.
A person handles the repeated physical work. The owner keeps the decisions: keep or discard, which box, and what needs special care.
I can be the last person to perform the task, but I must remain the final decision-maker.
Without that distinction, delegation becomes dumping.
If this works for packing, unpacking at the new home can be tested the same way. Expand only what proves useful.
4. A short, simple job is easier to accept than expected
The offer matched the structure of short one-off work.
A Japanese survey published in 2026 found leading reasons for starting short jobs included earning spending money (33.2%), choosing convenient free time (32.9%), working in a preferred place (24.1%), and starting without an interview (22.8%).[4]
A separate national survey found that one job lasted 3.5 hours on average. Work lasting two to under three hours accounted for 15.9%.[5]
The listing had the key features:
Short. Simple. Flexible. No interview. ¥3,000 at the end.
For the requester, ¥3,000 was roughly the cost of one meal out. Paying that amount to remove hours of tedious work felt worthwhile.
Safety still matters. Entering a stranger’s home can feel risky. That may partly explain why the applicants were all men.
A responsible listing should use daytime hours, explain the work clearly, separate valuables, and allow a public meeting point or direct arrival.
5. Sometimes a direct answer matters more than nearly 100 five-star reviews
One candidate had nearly 100 strong reviews. That was a major positive signal.
But when given two transportation options, the person replied only, “I live pretty far away.”
That explained the problem but did not move the job forward.
A useful reply would say:
- “I can reach an intermediate station; please pick me up there.”
- “I can travel myself if transportation is covered.”
- “If pickup is difficult, I’ll pass this time.”
A good reply includes the problem, a proposal, and the action the person will take.
A constraint alone creates more questions: What do you want? How far can you come? What about the return trip?
The selection criteria therefore became:
ratings, distance, reply speed, direct answers, and whether the person offers a workable next step.
Research on personnel selection also finds that job-related structured questions predict performance better than an unplanned chat.[7]
A few messages cannot reveal someone’s whole personality. They can, however, show how many extra rounds of coordination this particular task may require.
6. Delegating the work does not remove the planning
Once a person is involved, new decisions appear:
- date and time;
- pickup point;
- travel cost;
- what may be touched;
- what counts as finished;
- who decides when something is unclear.
Physical work falls, but coordination remains.
Too much instruction can take as long as doing the job yourself. Too little instruction makes the helper stop and ask.
The same rule helps when you are the customer. Do not stop at “It doesn’t work.” Say:
“I tried A and B. I want C. I can also do D.”
That lets the other side decide without another interview.
Heavy AI use makes this extra communication turn easy to notice. You repeatedly see the difference between an answer that advances the task and one that needs another explanation.
7. Keep working while a reply is pending
A nearby candidate read the message but did not reply for about thirty minutes.
Thirty minutes was not slow. The mistake would be stopping all packing until the reply arrived.
So the boxes were built and packing started.
If the person confirmed, they could join halfway through. If not, the work would continue.
Now, their reply time was not the same as my stopped time.
Good delegation is not about waiting better. It is about keeping the rest of the work moving while one part is uncertain.
8. The same skill matters in marriage, but looks and hobbies still matter
Shared life contains a great deal of invisible planning:
- remembering dates;
- researching services;
- making reservations;
- noticing broken items;
- planning trips;
- checking money and contracts.
A 2026 study of 235 cohabiting couples examined physical, mental, and emotional household work. Larger differences in how partners viewed the division of work were associated with poorer relationship quality.[8]
So long-term compatibility includes one practical question: Can we run daily life together?
That does not make attraction or hobbies unimportant.
Looks matter. Shared fun matters. Hobbies may matter. It also matters that one person does not spend a lifetime issuing every household task.
9. Conclusion: Ask “Do I need to do this?” before “Can I do this?”
The old order was:
I do it first. If it becomes painful, I ask a person. Maybe I use AI.
The new order is:
Can AI do it? Can a person do it? I execute last.
This does not mean disappearing from the work. Someone still defines the goal, chooses the person, checks the result, and keeps the important decisions.
AI has no arms. Humans do not behave like perfectly written instructions. Sometimes doing it yourself is fastest.
The useful skill is not outsourcing everything. It is choosing whether AI, another person, or you should perform each part.
If heavy AI use teaches only clever prompting, not one cardboard box gets packed.
If it teaches how to divide work while keeping judgment, AI can finally reach into the physical world.
Those arms usually belong to the guy who arrived for ¥3,000.
References
[1] Tomašev, N., Franklin, M., & Osindero, S. (2026). Intelligent AI Delegation. https://arxiv.org/abs/2602.11865
[2] Fábrega, J. (2026). Governing Delegation to Generative Artificial Intelligence: Human Direction, Work-Related Orientation, and Modes of Use. https://arxiv.org/abs/2608.17624
[3] Chandler, G. N., DeTienne, D. R., McKelvie, A., & Mumford, T. V. (2011). Causation and effectuation processes: A validation study. https://doi.org/10.1016/j.jbusvent.2009.10.006
[4] Recruit Works Institute (2026). Short-term spot-work motives. https://www.works-i.com/research/project/spotwork/deta/detail007.html
[5] NetAsia (2025). Spot Work Survey 2025. https://netasia.co.jp/report/4881/
[6] Kirei One. Packing and unpacking service. https://kireione.co.jp/service/archives/182
[7] Sackett, P. R., Zhang, C., Berry, C. M., & Lievens, F. (2022). Revisiting meta-analytic estimates of validity in personnel selection. https://pubmed.ncbi.nlm.nih.gov/34968080/
[8] Coundouris, S. P., & Henry, J. D. (2026). A dyadic examination of the division of household labor and relationship quality from a tri-load perspective. https://doi.org/10.1038/s41598-026-60727-z
