Lesson 6: Ask for Outcomes, Not Just Artifacts
Lesson overview
Most requests to AI describe a deliverable — a document, a summary, an email. The better question is: what should this deliverable actually accomplish? This lesson teaches the difference between asking for a thing and asking for a result, and shows why outcome-oriented requests produce more useful output.
What this means
An artifact is the thing you are creating. An outcome is what that thing is supposed to do.
"Write a summary" is an artifact request. "Write a summary that helps the leadership team decide whether to move forward with the partnership" is an outcome request. Same general task — very different direction.
When AI knows the intended outcome, it can make better decisions about what to include, what to emphasize, how to structure the output, and what level of detail is actually needed. When all it knows is the artifact, it produces a generic version of that artifact.
Why it matters
Deliverables without outcomes tend to be complete but not useful. They check the box of existing without achieving the purpose they were built for.
We have all read meeting summaries that could not lead to a decision, proposal documents that did not move anyone, or blog posts that did not say anything. Those items were technically complete. They just did not have a clear sense of what they were supposed to accomplish.
Defining the outcome before you ask for the artifact closes that gap.
What most people do wrong
Describing the format without describing the function
"Write a product brief." A product brief is a format. What does this one need to do? Get a feature approved? Align stakeholders on scope? Onboard a new engineer? Each of those purposes produces a different document.
Confusing activity with impact
"Summarize the research" is an activity. "Summarize the research so the team can prioritize which features to build next" is a result. The summary will be organized and weighted differently depending on which one you ask for.
Treating all deliverables as generic
A memo is not just a memo. An email is not just an email. Each has a purpose in a real context. Naming that purpose gives AI the information it needs to serve it.
The artifact vs. outcome comparison
| Artifact request | Outcome-oriented request |
|---|---|
| Write a meeting summary | Write a meeting summary that captures decisions made and the action items each person is responsible for |
| Write a case study | Write a case study that helps prospects in logistics see themselves in the problem we solved |
| Write a proposal | Write a proposal that gets budget approved for a $15k tooling investment — focused on ROI, not features |
| Write a blog post | Write a blog post that drives newsletter sign-ups from readers who manage distributed teams |
| Write a job description | Write a job description that attracts experienced candidates while discouraging early-career applicants who will not advance past the interview |
The right column does not just name a different outcome — it fundamentally changes what the artifact should contain.
Weak example
Write a proposal for the new internal dashboard project.
What can go wrong: AI writes a standard project proposal. It is probably structured fine. But it does not know whether the goal is to get budget, to align stakeholders, to document the scope, or to get executive sign-off. Different purposes require different emphases, different language, and different calls to action.
Strong example
Write a project proposal for a new internal dashboard. The purpose is to get approval from the VP of Operations to greenlight the build. She cares primarily about time-to-value and implementation risk. The project would reduce manual reporting by 6 hours per week for two teams. Frame the proposal around ROI and risk mitigation. Keep it under one page. Tone: concise and credible — not sales-y.
What is better: The outcome is named (get approval from a specific person). Her decision criteria are identified. The proposal is framed accordingly.
Practical exercise
Convert these artifact-only prompts into outcome-oriented requests. For each one, think: what is this piece of writing supposed to accomplish? Who needs to act on it or be moved by it?
- "Write an announcement email about the new product feature."
- "Write a project retrospective."
- "Write a summary of the competitor analysis."
- "Write a training document for new hires."
For each rewrite, add: who the reader is, what decision or action the output should support, and any framing that focuses the content on that outcome.
Reflection prompt
- Think about the last piece of work you created with AI. What was it supposed to accomplish? Did the output actually serve that purpose?
- Is there a type of document you regularly produce where the outcome is unclear even to you? What would clarifying it change?
- What does "useful" mean for the three most common things you create at work?
Key takeaway
Artifacts are containers. Outcomes are what they are for. Tell AI what the thing is supposed to accomplish — not just what type of thing it is — and the output will be shaped around something real rather than something generic.