Synthetic dataInvented manufacturer, documents and answers. No model runs on this page.
Expert Test Round
An agent has been set up on one floor-scrubber manufacturer's documents. Before launch, three of the manufacturer's experts asked it 30 real-world questions and marked each draft answer. This page sorts every miss to the person who can fix it and checks the go-live gates.
Result
Summary
Go-live gates
Open items by owner
Mark an item resolved to see how the gates change. A resolved item counts as passing until the next round reruns its questions.
Docs ask for the manufacturer
One grouped message to the manufacturer's champion, written from whatever is still open on their side.
All 30 questions
ID
Asked by
Type
Question
Verdict
Select a row to see the draft answer, its citation, the expert's note and the rule applied. The same 30 questions are rerun after each round of fixes, so a fix that breaks another answer shows up.
The fixed rules
Written down before the round ran. Each kind of miss goes to one fix and one owner.
Miss
When
Fix
Owner
Documents the agent was set up on
Assumptions
Pass means the expert accepted the answer as written. A warranty or safety question only passes when the agent hands it to a person.
The agent's answers are pre-written synthetic text. No model runs on this page.
The manufacturer, models, part numbers, prices, documents and expert notes are invented.
The 90% bar for each expert is an assumption, not a known standard.
Due dates are assumptions, counted in days after the test round.
Marking an item resolved counts its questions as passing until the next round reruns them. Marks are kept only in your own browser.
What I'd need to learn
How do a manufacturer's experts send test questions today, and who on their side signs off on launch?
When the parts list and the price book disagree, who decides which is current?
Is the launch bar a pass rate, a set of must-pass questions, or the expert's call?
Which misses get fixed through configuration, and which go to engineering?