Chapter C07 · Interactive Course
Give a useful brief
Supply what a task needs, without relying on prompt tricks.
Core Learning Concept
Build a brief that another person could follow without knowing what is in your head. Specify the result, reader, relevant background, format, and boundaries. Then compare the prepared response with those requirements. A helpful tone is welcome, but it cannot compensate for a missing date or an unusable plan.
What You’ll Learn in This Chapter
-
Name the outcome
Turn a vague topic into a clear requested result.
An outcome is the result you want to have when the task is finished. “Reading habits” names a topic; “Create a four-week reading plan with one daily activity” names a deliverable. A specific result helps both the assistant and you recognize whether the response actually completes the task.
-
Say who it is for
Specify the intended audience's needs.
The audience is the person who will use the result. Describe what they already know and what they need to do. A beginner may need a term explained and a first step shown. Asking the assistant to act like an expert can shape a reply, but it does not create professional qualifications.
-
Include relevant background
Select context that materially affects the task.
Relevant background changes what a good answer would be. Available time changes a study plan; an unrelated birthday usually does not. Before including a fact, ask whether leaving it out could make the result unsuitable. This keeps the brief focused and reduces unnecessary sharing of personal information.
-
Choose the output format
Request a structure that makes the result usable.
Choose a format that makes the result easy to use. A list exposes the order of steps; a table aligns several options against the same criteria; a paragraph explains a judgment. In the examples, compare how quickly you can find the information you need in each format.
-
State the boundaries
Specify one limit that the output must respect.
A boundary is a limit that a useful answer must respect. “Use materials we already have” prevents a plan that requires purchases. State the limit explicitly and make it checkable. Something obvious to you may never have reached the assistant, so it cannot reliably shape the answer.
-
Show one good example
Use an example to demonstrate the desired style or structure.
An example shows the pattern you want more concretely than an adjective such as “professional.” Supply a short sample and explain which parts to imitate: structure, tone, or level of detail. Mark facts that must be replaced so the assistant does not copy an old date into a new notice.
-
Define what good means
Choose a simple criterion for judging the output.
A success criterion is a check you can perform on the finished result. “Includes the date, time, and place” is easier to assess than “looks good.” Write a few criteria before asking, then inspect the draft yourself. Self-review by the same assistant can help find issues, but is not independent evidence.
-
Build a brief for someone else
Assemble a reading-plan brief with a deliverable, audience, relevant background, format and one boundary, then judge an authored answer against it.
In-Depth Practice & Real-World Application
Workplace Scenario: Briefing an AI model on a complex task without relying on vague "magic prompt" tricks.
Worked Example: The 4-Part Brief Architecture
Common mistake: “Write a memo about our new remote work policy.” (Lacks tone, requirements, audience, and key policy changes)
Effective pattern: “Role: Operations Director. Audience: 50 remote engineering staff. Task: Announce updated core collaboration hours (10am–3pm EST). Constraints: Keep under 200 words, warm tone, clear bullet points for next steps.”
The 4-part brief (Role, Context, Goal, Constraints) supplies everything the model’s neural probability needs to produce publication-ready drafts.
Key Takeaways
- Prompt engineering is simply clear professional briefing, not secret code words.
- Constraints (length, tone, excluded topics) are just as important as the primary request.
- Providing an intended audience prevents the AI from defaulting to robotic corporate jargon.
Workplace tip: Use the R-T-C-C framework: Role (who AI speaks as), Task (what to do), Context (background details), Constraints (length, exclusions, format).