GOOD COMPANY PROJECT
Small experiments. Honest results.
Working notes on practical human–AI collaboration: the problem, each contribution, the checks, and what remains uncertain.
The example below is fictional and untested. It illustrates a method; it is not a report of a real result.
A clearer volunteer handoff
Experiment 01 · Fictional teaching example · No real-world result claimed
The starting message
“Could someone sort the supplies before people arrive Saturday? The list should be in the cupboard. Let me know if anything’s missing.”
The volunteer still has to guess who owns the task, which room to use, when to finish, and where to ask for help.
A prompt you can reuse
“Review this handoff for missing details. Ask about the person responsible, place, finish time, checklist, meaning of done and contact for problems. Don’t invent answers. Then write a short handoff using only facts I confirm.”
The person supplies the facts
In this made-up scenario, the coordinator confirms Room 2, Saturday at 9:30 a.m., a checklist in a blue folder beside the cupboard, and supplies to be placed on the two front tables. Sam is the proposed volunteer; Lee handles questions in the planning chat.
The revised message
“Hi Sam, can you set out the supplies in Room 2 by 9:30 a.m. Saturday? The checklist is in the blue folder beside the cupboard. Put the listed items on the two front tables. If anything is missing or the timing won’t work, message Lee in the planning chat. Please reply to confirm you can take this on.”
Who does what?
The person supplies the facts, checks that the supplies exist, and gets the volunteer’s agreement.
AI flags gaps and organizes confirmed details. It cannot check the cupboard or know whether Sam is available.
Check whether it helps
Before the task, ask the volunteer to explain the plan in their own words. Fix anything they still have to guess. Afterward ask: “Was that enough to get started? What was still missing?” Record the answer and corrections before calling the experiment useful.
Result status: untested illustration. Clearer wording alone does not show that a real handoff improved. Note errors, extra work, and unresolved questions alongside any measured benefit.
Try the prompt with invented or non-sensitive details. Compare the draft with the facts before sending anything.