Purpose
Use this lesson to separate research, reasoning, drafting, and review into controllable stages. The aim is to produce work that a colleague can review and act on, not merely text that sounds convincing.
Lesson Explanation
Scenario: Prepare a recommendation on remote-work policy.
Better instruction: First list the decision criteria. Then compare the provided options against them. Finally draft a recommendation with assumptions and risks.
A large prompt can hide errors because research, analysis, and writing blur together. Stage the work so that you can inspect the criteria before trusting the final recommendation. Do not ask for hidden reasoning; ask for visible, checkable intermediate artifacts. The output exposes criteria, comparison, assumptions, risks, and a recommendation.
Work in a short cycle: provide the relevant input, state the result and limits, inspect the draft against the acceptance criteria, then revise the instruction when a requirement is missing. Keep the final human decision with the person responsible for the work.
Practice Questions
-
A single prompt asks an AI to “research, analyze, and recommend” a solution to a complex problem all at once. What does this lesson say can happen when this is requested as one large step?
View Answer
Errors can hide because research, analysis, and writing blur together, making it hard to inspect any individual part before trusting the final result.
-
Rewrite a request to “recommend whether we should expand into a new market” by breaking it into visible stages, per this lesson’s approach.
View Answer
Something like: “First, list the criteria for evaluating this market. Then, compare the market against those criteria using the provided data. Finally, draft a recommendation with assumptions and risks stated separately.”
-
What does this lesson recommend requesting instead of “hidden reasoning”?
View Answer
Visible, checkable intermediate artifacts.
-
A complex prompt produces a single polished recommendation with no visible intermediate steps to review. What risk does this lesson identify with this approach?
View Answer
It becomes hard to catch an error in the underlying reasoning or criteria, since only the final polished output is visible.
-
A manager wants to evaluate three vendors and asks the AI to just “tell me which one to pick.” How might this lesson’s approach improve this request?
View Answer
Breaking it into stages – first listing evaluation criteria, then comparing vendors against those criteria, then drafting a recommendation – so each part can be checked.
-
Why does staging complex work allow you to “inspect the criteria before trusting the final recommendation,” according to this lesson?
View Answer
Because each stage produces a visible, checkable output, letting you verify the criteria are reasonable before the final recommendation is built on top of them.
-
A prompt asks the AI to first identify assumptions, then build an analysis based only on stated assumptions, then draft a conclusion. What principle does this reflect?
View Answer
Breaking down complex work into visible, staged steps rather than requesting a single large, opaque output.
-
A large one-shot prompt for a complex financial recommendation produces a confident-sounding answer, but a reviewer later discovers the underlying assumptions were flawed. What does this lesson suggest could have caught this earlier?
View Answer
Staging the work so the assumptions were a visible, separate step that could be reviewed before the final recommendation was generated.
-
What kind of prompt structure does this lesson recommend for genuinely complex tasks?
View Answer
A staged structure where research, comparison, and recommendation (or similar stages) are requested and reviewed separately, rather than blurred into one large output.
-
A prompt asks the AI to compare two options against defined criteria as a distinct, visible step, before drafting any final recommendation. What is the purpose of this staging, per this lesson?
View Answer
To allow inspection of the comparison itself before trusting whatever recommendation is built from it.
-
Why might “ask for the answer directly” be a weaker approach for complex work than staging it into visible steps, according to this lesson?
View Answer
A direct answer to a complex question can hide the reasoning and potential errors within a single blurred output, while staged steps expose each part for review.
-
A request breaks a complex hiring decision into: list the role requirements, compare candidates against those requirements, then draft a recommendation with reasoning. What does this reflect?
View Answer
The staged, visible-artifact approach to breaking down complex work that this lesson recommends.