Find the right work for AI
Want more background?
The idea in everyday life
Map the task before choosing the tool. Good early candidates are frequent, reversible drafts you can check against permitted source material. The four supplied profiles let you practice without a job or private records. Treat each profile's sensitivity and permission rules as real constraints; a larger workload does not override them.
AI generates responses from patterns it has learned. It can help rewrite or organize information, but you must check whether its answers are true.
Four fictional tasks to compare
| Task | Frequency | Manual time per run | Data and permission | Consequence of an error | Checking and reversibility |
|---|---|---|---|---|---|
| A — Draft public study-club meeting notes | 1 per week | 30 minutes | Low sensitivity: public notes; reuse with AI is permitted. | A wrong action item could mislead someone if the draft is used unchecked. | Check every action, owner, and date against the notes; edit before use. You are the practice reviewer. |
| B — Summarize private support requests | 5 per week | 20 minutes | High sensitivity: names, contact and health details; AI use is not approved. | Private information could be exposed or an urgent request missed. | Understanding may require confidential context. No approved tool or reviewer is available. |
| C — Turn your own study notes into a checklist | 4 per week | 15 minutes | Low sensitivity: your original notes with no private details. | An omission or misstatement could leave a study gap. | Compare each item with notes you understand; revise the draft before using it. You review it. |
| D — Make final volunteer-selection decisions | 2 per week | 45 minutes | High sensitivity: private applications. | An unfair decision could exclude a person from an opportunity. | A qualified human must make the final decision; a fluent AI explanation does not establish fairness. |
One step at a time.
Use the sample or information you have permission to use. Keep your work as you go. You can pause between steps.
- 1
Use the four fictional task profiles above. You do not need a job, private records, or an AI account for this comparison. Rate sensitivity and permission as if the described inputs were real; do not collect those records. Mark each task suitable for a reviewed draft or unsuitable as stated. Frequency and time cost cannot cancel missing permission, serious harm, or poor checking.
- 2
Read the A/D worked comparison below. Identify the permitted input, draft output, reviewer, and reason for keeping the result reversible. Keep final authority with a qualified human for legal, medical, financial, safety, hiring, or other high-impact decisions.
- 3
Now compare C with B yourself. Choose the better first draft workflow and justify it using frequency, time, sensitivity, error consequence, and ease of checking. Explain why the other task is unsuitable as stated.
- 4
Rank the suitable tasks, mark the others excluded for now, and select exactly one for later testing. Name its audience, output, permitted inputs, reviewer, and one observable acceptance check. You can be the reviewer for this fictional exercise. Do not count baseline workload as time saved; a later trial must include checking and correction time.
See how it works.
A takes 1 × 30 = 30 manual minutes per week; D takes 2 × 45 = 90. Choose A for a first test despite D's larger time cost. Its permitted public notes can produce a draft action list for the study club. You check every action, owner, and date against the source and mark missing details as not specified before anyone uses it. That is an observable acceptance check, and the draft can still be corrected. D makes a consequential final selection: a qualified human keeps that decision. These are baseline workloads, not promised savings; later trials must include checking and correction time. After making your own C/B comparison, check your reasoning: C is suitable for a reviewed draft; B is excluded as stated because permission, data handling, and review are unresolved. C represents 4 × 15 = 60 manual minutes weekly, versus B's 5 × 20 = 100; the larger number does not make B acceptable. A and C are both eligible first tests. Rank them with a reason and choose one, keeping the other for later.
Did you cover these?
- My inventory compares all four tasks and excludes B and D as stated, with reasons tied to permission, potential harm, or final authority.
- I justify C over B, rank the suitable A/C drafts, and select one with a named reviewer, permitted input, output, and an observable acceptance check.
- I distinguish baseline weekly workload from measured savings; my planned trial includes checking and correction time.
What to finish
A ranked fictional task inventory with 1 low-risk workflow selected and justified for later testing.
Watch for this
Choosing a flashy task that is rare, hard to verify, or too risky to delegate.
How can I tell I understand it?
You can explain why each chosen workflow is valuable, checkable, and allowed.
Quick knowledge check
Answer all 3 questions. Get at least 2 right and the lesson completes automatically.
You’ll get an explanation for each answer. Signed in? Your results save to your account.
Pass the check. We’ll save the progress.
Answer all 3 questions above and get at least 2 right. You can review the explanations and try again.
Go to the knowledge checkOptional course overview video
AI and Machine Learning with Google Cloud
From Google Cloud. Watch here or open it on YouTube .
Use the course video's examples to separate generative drafting from authoritative decision-making; select a task where you already understand the desired result.Tools, templates, and lesson sources
Tools and templates
Tools open in a new tab so you can keep your place here.
Sources and further reading
Explore the references behind this lesson.
Report a problem with this lesson
Tell us what went wrong. Please leave out personal information and private project contents.