Applied Checkpoint: Choose the Right Method
Complete the activity before revealing the model answer. Record one change you would make after comparison.
Instructions
For each situation, record:
- Main task pattern.
- Evidence requirement and unacceptable error.
- Input classification.
- Recommended method: AI, another tool, controlled combination, or human-led work.
- Human authority and verification.
- Stop condition.
Use the cautious course labels for data. In real work, follow the organisation’s policy.
Situation 1: public event announcement
A community centre has approved public facts for an event. It wants a friendly 100-word announcement. A communications officer will check it before posting.
Situation 2: expense total
An employee has twelve approved receipt amounts and needs the exact total in the organisation’s currency. The result will be submitted for reimbursement.
Situation 3: customer complaint themes
A team wants to paste 500 complaints containing names, phone numbers, account details, and free-text descriptions into a free public AI chatbot to find common themes.
Situation 4: current application deadline
A learner needs the closing date for a government scholarship. The date may have changed since last year.
Situation 5: hiring decision
A manager wants an AI assistant to rank applicants and automatically reject the bottom half. The applications contain personal employment and education details. No bias test, appeal process, or authorised review has been defined.
Reveal the model answer
Situation 1
- Pattern: generation from supplied facts.
- Evidence: every factual claim must match the approved event information.
- Data: public.
- Method: one AI draft is reasonable, or a person can write it directly.
- Authority and check: communications officer compares the draft with the facts and approves posting.
- Stop: remove or query any added date, person, promise, price, or contact detail.
Situation 2
- Pattern: calculation.
- Evidence: the twelve approved amounts and the organisation’s reimbursement rules.
- Data: likely internal; receipts may also contain personal or financial information.
- Method: approved spreadsheet, calculator, or expense system. AI is not needed for the total.
- Authority and check: employee checks every entered value; finance process approves reimbursement.
- Stop: pause for missing, unreadable, duplicate, or unusual receipts.
Situation 3
- Pattern: classification and analysis.
- Evidence: the complaint set and an agreed theme method.
- Data: personal and confidential; some details may be especially sensitive.
- Method: do not use the proposed public chatbot. Use an approved protected process with minimised or safely prepared data, or keep the work human-led.
- Authority and check: data owner, privacy/security contacts, and an authorised analyst define the permitted method.
- Stop: stop before upload because tool approval and data handling are not established.
Situation 4
- Pattern: search.
- Evidence: the current official scholarship page or notice.
- Data: public.
- Method: find and check the official source. AI may explain the notice but is not the evidence.
- Authority and check: learner checks the date, year, eligibility group, time zone, and update date.
- Stop: pause if official pages disagree or the date is missing; contact the responsible office.
Situation 5
- Pattern: high-impact decision.
- Evidence: authorised hiring criteria, complete relevant evidence, and a fair approved process.
- Data: personal and confidential.
- Method: reject the proposal. Keep the decision human-led under an approved hiring process. Any future AI support requires legal, privacy, fairness, security, and operational review.
- Authority and check: authorised hiring team retains responsibility and provides appropriate review and challenge routes.
- Stop: stop before applicant data is uploaded or any rejection is automated.
Reflection
Which situation first looked suitable for AI but changed after you considered evidence, data, or authority? Write one sentence explaining what changed your choice.