Task Suitability Checklist
Decide whether AI, fixed software, another method, or human-led work is the right choice.
For AI Foundations learners · Printable PDF · 2 pages · Also available online below
Use this checklist before giving real work or information to an AI system.
T - Task
- What action and result are required?
- Is the work generation, transformation, extraction, classification, analysis, search, calculation, or decision?
- Does a mixed task need to be split into smaller parts?
A - Accuracy and authority
- What evidence must support the result?
- Does the evidence need to be current or official?
- Which errors are unacceptable?
- Who or what could be harmed by a mistake?
- Can the result or action be reversed?
- Who has authority to review and approve it?
S - Sensitivity
- Is each input public, internal, confidential, personal, or prohibited?
- Is the proposed tool approved for that category?
- Can names, secrets, and unnecessary details be removed?
- Do I know how the service stores, uses, shares, and deletes inputs?
K - Kind of tool and control
- Would search, a calculator, a spreadsheet, fixed software, or a specialist be better?
- Is one reversible AI draft enough?
- Does the task need a controlled multi-step workflow?
- If tools or actions are delegated, are permissions, limits, approvals, monitoring, and logs defined?
- Is human-led work the safer choice?
S - Stop conditions
- Which missing evidence should stop the task?
- What happens if trusted sources disagree?
- What happens if sensitive information appears?
- Which unusual result or irreversible action requires escalation?
- Who receives the task after it stops?
Recommendation sentence
Use
[method]for[bounded task]because[evidence and risk reason]. Use only[permitted input].[reviewer]must verify[checks]and approve[action]. Stop and escalate to[owner]if[conditions].
