Illustrative experiment: offer a bounded proposal-preparation pilot for independent consultants. This is a teaching scenario, not a new service listing or an earnings claim. The existing income guide supplies a hypothetical 14-day test and a cost example.
Open the detailed experiment and cost example →
Step 1 · Human + research worker
Find a reachable buyer
List skills you can deliver, audiences you can reach and recurring problems you have observed. Read relevant public sources and conduct voluntary customer conversations.
Leave with: A shortlist with evidence, sources and access routes.
Review before continuing: Is the problem based on behavior, or only an AI-generated persona?
Step 2 · Product worker
Choose the risky assumption
Separate whether the buyer cares, will pay, can be reached and can be served sustainably. Choose the uncertainty that could invalidate the idea first.
Leave with: One hypothesis and the evidence that would disprove it.
Review before continuing: Do you need an interview, prototype, paid pilot or delivery test? Do not build software to answer a buying question.
Step 3 · Designer + delivery worker
Make a small sample
Create one realistic sample with fictional or authorized data. Describe the deliverable, exclusions, price hypothesis and acceptance criteria.
Leave with: A sample and a plain-language offer.
Review before continuing: Can you deliver the promised outcome, including review and corrections, within the effort limit?
Step 4 · Marketing worker + human
Run one distribution test
Choose one appropriate channel. Draft an invitation or useful demonstration with a clear next step. Personally review publication and outreach. Record the audience and source.
Leave with: Approved drafts and a record of relevant invitations or visits.
Review before continuing: Respect declines and channel rules. Likes are not purchases; absent traffic does not disprove demand.
Step 5 · Delivery + QA worker
Deliver and count the costs
If a buyer agrees, fulfill the defined pilot and obtain acceptance. Record all setup, selling, delivery and review time, plus direct costs and refunds.
Leave with: Accepted work and a complete evidence ledger.
Review before continuing: Include failed attempts. Do not label revenue as profit or ignore your time.
Step 6 · Analyst + human decision-maker
Continue, change or stop
Compare observations with your preselected decision rule. Explain missing data and alternative explanations. Select the next smallest test, or stop at the agreed limit.
Leave with: A decision memo and next review date.
Review before continuing: A few purchases are directional evidence, not a stable conversion rate or proof of repeatable wealth.
Copy the starter prompt
Evaluate this income idea using only my skills, customer notes and sources. Separate facts, assumptions and unknowns. Identify the riskiest assumption and propose one small test with a cash limit, total-hour limit, primary metric, denominator and continue/change/stop criteria. Do not invent demand, market size, customer quotes or expected earnings.
Where this workflow can break
Agents may confidently generate market estimates or recommend building a full product. Ask for source-linked evidence and a disconfirming test. A rejection can improve the offer; a missing audience means the distribution test needs work.
Your practice assignment
Complete one experiment card and one sample. Set the limits before reaching buyers. Record the outcome honestly, including a decision to pause.
Get the brief, handoff and review templates
Keep a useful result log
Record what you attempted, what you accepted, corrections, cash spent, total human time and the next decision. A published output or completed lesson is an artifact; demand, learner performance and business results require separate evidence.
Explore the other workflows
Build and publish a website with agents · Create a useful course with an agent team