Prompting and Output Verification
A working session on getting reliable results and, more importantly, on checking them. Built around the verification habits that separate occasional AI users from dependable ones.
What participants can do afterwards
- Structure a prompt with role, context, constraints, and output format
- Iterate on a weak result rather than abandoning it
- Apply a repeatable verification routine to AI output before it leaves your desk
- Document a prompt so a colleague can reuse it
Modules
3 modules, 7 facilitated hours.
1.Structuring a request
2 hrsRole, context, constraints, format. Practised against weak prompts the group brings in.
- Rewrite a weak prompt into a structured one
2.Iteration instead of abandonment
2 hrsWhat to do with a poor first answer, and how to tell a fixable result from a dead end.
- Recover a usable result from a poor first answer
3.Verification routines
3 hrsBuilding a checking habit proportionate to the stakes of the work.
- Apply a verification routine before releasing AI-assisted work
How it is delivered
A facilitator runs the course live, in person or live online. Attendance is taken per session, participants submit work against the modules above, and a facilitator grades it. Those records stay with the organisation and can be exported.
Further reading
An answer nobody checks is a liability is the public briefing that sits next to this course.