Chapter 1 · Lesson 1
Define the Job Before You Prompt
Turn a loose request into a clear job with an outcome, audience, and source material.
You want AI to help with a project update, but “write an update” could mean a dozen different things.
Start with the decision or action
A useful prompt begins with what the finished work must help someone do. “Summarise this project” describes an activity. “Help the steering group decide whether to delay launch” describes the job. The second version gives every later choice a purpose.
Name the audience and evidence
The same source material should become a different answer for an executive, a customer, or a technical team. State who will use the result and identify the material the answer must rely on. This reduces generic filler and unsupported invention.
Separate inputs from instructions
Label source text clearly instead of mixing it into the request. A simple structure such as Goal, Audience, Source material, and Deliverable makes the prompt easier to inspect, reuse, and correct when the answer misses the mark.
Build the mental model
A usable brief connects the request to a decision, an audience, and evidence.
Use a five-part job frame
Write Purpose, Audience, Inputs, Deliverable, and Done when. If one part is missing, the model must guess or produce work you cannot evaluate.
Test the brief before sending it
Ask whether a colleague could perform the task from the same brief. If not, the model probably cannot either. Fix the job definition before polishing the wording.
A project manager needs a steering-group update from meeting notes.
Write a professional project update from these notes.
Purpose: help the steering group decide whether to delay launch. Audience: senior leaders with no technical detail. Inputs: use only the meeting notes below. Deliverable: a 150-word update with Status, Evidence, Risk, and Decision needed. Done when: every claim is traceable to the notes and missing information is labelled.
The improved brief defines the decision, reader, evidence boundary, output shape, and quality test. The result can be reviewed against explicit conditions.
Why job framing matters beyond prompting
A vague prompt often reveals a vague process. Clarifying the purpose and done condition improves both AI-assisted and human work because everyone can see what the deliverable is for.
Do not add detail indiscriminately. Include information that changes the output or the way it will be judged. Background that does neither is noise.
Frame one task you expect to repeat
Choose a real task and complete all five fields. Then remove any sentence that would not change the result.
Purpose: [...] Audience: [...] Inputs: [...] Deliverable: [...] Done when: [...]prompt choice
Which request defines the job most clearly?
Compare the prompts closely. Look for useful context, constraints, and a clear output.
Define what the output must help someone do before choosing the words of the prompt.