Turn an open question into a deliverable you can review

Turn an open request into a work artifact that a person can review, correct, and use. The assistant organizes facts, makes assumptions explicit, and keeps remaining checks visible. It does not replace expert review or the interactive Quest flow, available to guests at /chat-assistant/guest and to signed-in users at /chat-assistant/app.

Three inputs to begin

1. Desired artifact: what do you need to produce, such as a plan, summary, table, or draft?
2. Decision owner or audience: who will review or use it?
3. Relevant facts: data, constraints, sources, and context that must be included.

When to use it

Use it when the initial request is ambiguous, notes need to become a concrete output, or you need a first artifact for review. It is useful for structuring meetings, preparing options, and identifying what is missing.

Do not use it to approve irreversible actions without review, invent facts, make decisions about people, enter secrets, or process information you are not authorized to share. When critical information is missing, ask for clarification or label the assumption.

Preparation worksheet

• Expected outcome and format.
• Audience, owner, and review date.
• Confirmed facts with their sources.
• Budget, timing, scope, or tone constraints.
• Allowed assumptions and unanswered questions.
• Criteria for a useful result.
• Data that must stay out of the chat.

How to use the assistant

1. Open AI assistant for guests or AI assistant when signed in.
2. Enter the three starting inputs: artifact, owner or audience, and relevant facts.
3. Add format, detail level, constraints, and intended-use date.
4. Ask it to separate facts, assumptions, and open questions.
5. Review the first output and correct context, priorities, or ambiguous terms.
6. Request a final version with decisions, open items, and sources clearly marked.
7. Run checks outside the assistant: figures, links, policies, permissions, and expert review.
8. Share only the sanitized artifact and retain the minimum context needed.

Worked small-business example

An online pet-supply store is preparing its monthly inventory meeting. The manager enters the desired artifact as “30-minute agenda with decisions and owners,” the audience as “manager and operations lead,” and relevant facts including four weeks of sales, three stockout items, and a supplier arrival date. She notes that the sales figures are preliminary. The assistant creates an agenda with a data-validation block, two replenishment options, and an owner table. The manager corrects one date, removes a recommendation based on an unconfirmed figure, and adds a cash limit. The final artifact lists decisions to make, assumptions used, and checks still required. It is shared internally without credentials, customer data, or banking information.

Share-card template

## Reviewable artifact

- Outcome: [what was produced]
- Supplied facts: [summary of confirmed data]
- Assumptions: [explicit assumptions]
- Final artifact: [sanitized link or content]
- Checks still required: [what must be verified]
- Review owner: [role, no personal data]
- Date: [YYYY-MM-DD]

Quality and safety checks

• Confirm that the artifact answers the original request.
• Distinguish facts, inferences, and generated wording.
• Verify numbers, dates, links, product names, and quotations.
• Review for bias, omissions, and unintended consequences.
• Do not paste passwords, keys, bank details, or unnecessary personal information.
• Confirm permissions before sharing documents or context.
• Keep a person accountable for approval.
• If a check has not been completed, leave it visible as pending.

Next step

If converting requests into artifacts is a recurring process, involves a team, handles substantial material, includes sensitive data, or has a meaningful review bottleneck, consider a business audit. Otherwise, keep using the Quest assistant for the next request and retain human review.