Worked research case
Protopilot
Try a product idea before handing it to engineering.
See how a product lead makes a small test for an idea they're considering.
Read-only worked research
- For
- Product teams using AI-assisted builders who need evidence before committing engineering time.
- Decision
- Teams turn untested assumptions into polished software too early, making it expensive to learn whether a real problem exists.
- Sessions
- 0 recorded
Author · The problem
The team has an idea to check.
A team can quickly make a polished screen. They still need to find out whether it solves the right problem before spending time on development.
Checkout confidence
Teams polish checkout ideas before testing whether buyers understand the commitment.
Author · What to find out
Will they make a test before starting development?
The creator expects product leads to make a tester flow before asking engineering to build. They write down what action would support that idea and what would count against it.
Checkout confidence
Teams polish checkout ideas before testing whether buyers understand the commitment.
After framing a problem and hypotheses, product leads will create a tester flow before asking engineering to build.
- Prediction
- At least four of five eligible product leads create the named tester flow without assistance before asking for an implementation.
- What would change your mind?
- Three or more eligible product leads skip the tester flow, require assistance, or ask to build first.
- Alternatives
- A lightweight assumption checklist may change behavior without an executable prototype composer.
- Confidence before testing
- 55% before participant evidence
Show authored assumption questions
- What commitment becomes expensive too early?
- A polished implementation makes the team defend a solution before it has tested the buyer problem.
Author · The prototypes
Try two ways to create the test.
Compare a guided composer with an assumption card. Both let the product lead create a flow for someone to try.
Checkout confidence
Teams polish checkout ideas before testing whether buyers understand the commitment.
Show model relationships and queries
- Commands
- createTesterFlow, requestBuildBeforeTest
- Entities and fields
ProjectDraft: title, problem, status
TesterFlow: projectId, name, createdBy
- Events
- testerFlowCreated, buildRequestedBeforeTest
- testerFlowProject
- TesterFlow.projectId → ProjectDraft (one)
- listProjectDrafts
- ProjectDraft
Show scenario definition
{
"records": {
"ProjectDraft": [
{
"id": "example-protopilot:scenario-record:ProjectDraft:1",
"title": "Checkout confidence",
"problem": "Teams polish checkout ideas before testing whether buyers understand the commitment.",
"status": "framed"
}
],
"TesterFlow": []
}
}Show authored model questions
- What observable behavior would reduce that risk?
- The product lead creates a runnable tester flow tied to the selected usage hypothesis before requesting engineering work.
Tester · Try the task
Take the product lead's place.
Choose the Checkout confidence project and give its tester flow a name. You can try this below.
Checkout confidence
Teams polish checkout ideas before testing whether buyers understand the commitment.
A product lead who currently uses an AI-assisted builder for product exploration.
Show screening and observation criteria
- Do you currently use an AI-assisted builder to explore product ideas? (eligibility)
Support signal: testerFlowCreated is recorded
Refute signal: buildRequestedBeforeTest is recorded
Show authored prototype-test questions
- What is the smallest honest test?
- Give an eligible product lead a framed checkout study and observe whether they can create a named tester flow without assistance.
Local simulation · not recorded
Author again · What happened
Check what happened.
The example run creates a tester flow. That matches the action the creator planned to look for, but they still need to test with real product leads.
Checkout confidence
Teams polish checkout ideas before testing whether buyers understand the commitment.
No participant evidence yet.
Show authored learning question
- What did participant sessions show?
- No participant sessions have run; the study is ready to test, but there is no outcome evidence yet.
The example command supports the authored observation rule. It is a local, runnable demonstration—not participant evidence.