Before you begin
Open the existing project so your instruction is grounded in the current app.
Write the instruction from the user’s perspective, then state the exact result you want to inspect.
Write the instruction
“Analyze this” is difficult to evaluate.
“Return a table with these six columns” is concrete.
Tell Empromptu AI what the user should receive
A table
A ranked list
A recommendation with reasons
A summary with citations
A completed form
A dashboard
A downloadable report
A proposed action that requires approval
Then name the fields.
Example
“Return a table with employee, requested access, recommended access, rule used, confidence, and reviewer decision. Put conflicts and low-confidence rows first.”
Now the output can be inspected. A user can correct it. An eval can test it. A downstream system can use it.
The best output is not the most impressive one. It is the one that makes the next human or system action obvious.
Common mistakes
Asking for a summary when the workflow needs structured fields.
Leaving required columns or fields implicit.
Not specifying what to show when information is missing.
Frequently asked questions
How long should a prompt be?
Long enough to define the user, job, input, rules, output, and important constraints. Clear structure matters more than word count.
Can I make several changes in one prompt?
Small related changes can be combined, but changes to different systems are easier to test and recover when requested separately.