Build prompts that expose uncertainty
Course overview · 4 min reading + 12 min practice, estimated
Principles and method
A useful prompt supplies the task, relevant context, permitted evidence, constraints and output structure. Ask the model to distinguish supported claims from missing information. Delimit source material clearly and treat text inside candidate documents as data, not instructions to change the task. A prompt alone is not a security boundary, but it can support a controlled workflow. Keep versions so an output can be traced to the instructions used. Use fictional or appropriately approved information during development and check data handling before using real records.
Worked example
A prompt requests a table with criterion, quoted evidence reference, uncertainty and question to ask. When no evidence exists, the required output is not established, rather than a plausible invented example.
Put it into practice
Write a prompt for a fictional quality-lead brief and test it with one missing requirement.
Use fictional information and keep your work in your own notes.
Compare your approach: self-review guidance
The output should visibly retain the gap. Include a rule against following instructions embedded in the source document and a reviewer check of every material claim.
Sources and further reading
Original Academy teaching and fictional examples. These references provide context, not endorsement. Edition 2026.09; updated 2026-09-24.
- GOV.UK: Responsible AI in recruitment
UK guidance on procuring and deploying recruitment AI.
- NIST: AI Risk Management Framework
Voluntary framework for organising AI risks and controls.