ChatGPT Task Briefs: Clarify a Vague Work Request
Turn a vague work request into a brief with confirmed facts, missing conditions and three useful questions. Check scope before starting the deliverable.
A clear request saves the wrong kind of work
“Make this presentable” may mean a report, a chart or a cleaned spreadsheet. If you start by guessing, a polished result can still answer the wrong question. ChatGPT can organize the instructions you have and expose gaps before you build the deliverable.
The aim is a checkable brief, not a plan approved by AI. Use one real request and its follow-ups. Separate your interpretation from the requester’s words so that a suggestion does not silently become a requirement.
1. Keep the request and context together
Copy the relevant request and follow-ups, with dates when instructions changed. Give each excerpt an ID such as R1. Add known audience, inputs and restrictions. A filename does not establish that its contents or access have been checked.
Remove personal details and unrelated confidential material. Use an approved tool and only the excerpts needed. Account data controls do not replace your organization’s permission to share information.
- Preserve the original wording of scope and deadline.
- Mark unconfirmed audience, reviewer and time zone as unknown.
- Label your interpretation as a proposal.
2. Practice with a small request
This fictional case describes a sales-summary request; it contains no actual sales figures. Several choices remain open even though there is enough information to begin a conversation.
“Two pages” limits length without specifying a file format. A Friday meeting does not establish the review time. Both gaps can change the work you prepare.
| Source | Information actually supplied |
|---|---|
| R1 | Prepare a sales summary for Friday’s team meeting. |
| R2 | The workbook contains April and May. |
| R3 | Keep the summary to two pages. |
| R4 | Workbook access is confirmed; the figures have not yet been verified. |
3. Separate facts from interpretations
Open the shared prompt and replace its bracketed fields with your request. Ask it to label conditions confirmed, assumed, missing or conflicting. Read the source IDs yourself: a label is useful only when it points to the right words.
The team meeting and two-page limit are confirmed. “Compare April with May” is a possible interpretation, not an agreed instruction. Purpose, reporting period and review arrangements still need clarification.
Open, edit and copy the full prompt
4. Ask questions that unlock decisions
Prioritize questions by their effect on the next action. The reporting period determines which rows to use. The decision the meeting should support guides your choice of content. Format and review timing shape the handoff.
A confirmation message could say: “I have the April–May workbook and two-page limit. Please confirm the reporting period, decision to support, and format and review time. I will treat the April–May comparison as a proposal until confirmed.” Review it before sending.
| Confirm | Why it changes the work |
|---|---|
| Reporting period | Select the right records instead of defaulting to the newest month. |
| Decision to support | Choose relevant comparisons rather than filling pages with available data. |
| Format and review time | Prepare a draft without inventing a deadline. |
5. Prepare what can be prepared
While awaiting answers, inspect the workbook within existing permissions. List sheets and columns, missing values and totals to check. You can prepare a blank brief without claiming that any analysis is approved.
Avoid distributing a finished report based on an unconfirmed interpretation. Save a dated follow-up as a new source when scope changes. Surface conflicting instructions rather than letting AI quietly choose between them.
6. Check the brief before you begin
Compare the brief with the source pack and confirm that the requester’s answer resolves the actual gap. “As discussed” is not evidence of a discussion nobody recorded. Add a short note of the confirmed decision.
You should be able to state the output, its purpose, what is outside scope and who reviews it. Any remaining uncertainty should stay visible.
- Does each confirmed condition have a matching source?
- Are assumptions separate from approved requirements?
- Are dates, times and time zones preserved?
- Do unanswered questions affect whether you can start?
- Have you reviewed the confirmation message yourself?
Related guides and sources
OpenAI notes that ChatGPT can be wrong. Check important claims against your source. This clarification workflow and fictional case are Praclane’s editorial method.
Photo: Nick Morrison / Unsplash.
