AI literacy does not mean knowing every new tool. It means knowing where AI can help, where it can fail and when a human must take responsibility.

The International Labour Organization's research suggests that task transformation is more likely than complete automation for most occupations exposed to generative AI. That is not a promise that every job is safe. It means a practical response is to learn how work is being reorganised rather than trying to guess which headline will be correct.

A useful worker does more than produce a fast first draft. They understand the goal, check the result, protect sensitive information and make a responsible decision.

Week 1: map your real tasks

Write down ten tasks you perform, have performed or expect to perform in your target role. Label each task:

  • Draft: producing an email, summary, checklist or first version.
  • Decide: choosing between options or accepting risk.
  • Relate: understanding a customer, colleague, patient or candidate.
  • Verify: checking facts, calculations, policy or quality.
  • Protect: handling private, confidential or safety-sensitive information.

AI is usually easier to test on low-risk drafting and organisation. It should not silently take over decisions, relationships, verification or accountability.

Choose one repetitive, low-risk task. Use public or invented information during practice. Do not paste identity numbers, medical details, customer records, confidential documents or unpublished employer data into a public AI service.

Week 2: practise a repeatable method

Use a simple four-part instruction:

  1. Context: explain the situation and intended reader.
  2. Output: state what you need, such as a checklist or short email.
  3. Constraints: set the tone, length and facts that must not be invented.
  4. Check: ask for uncertainties and items requiring human verification.

Generate a first draft, then compare it with work you would trust. Mark every unsupported claim, vague phrase and missing detail.

Repeat the exercise with three workplace examples:

  • Turn rough meeting notes into an action list, then verify every owner and date.
  • Draft a customer response, then check it against the real policy.
  • Convert a job advert into an interview-preparation checklist without inventing company information.

The goal is not to become impressed by fluent text. It is to become good at detecting when fluent text is wrong.

Week 3: add the human layer

AI can make average-looking output cheap. Your advantage comes from the work around the output.

Practise four abilities:

  • Judgement: decide which recommendation fits the situation.
  • Explanation: translate technical information into language another person can use.
  • Empathy: notice concerns or context that a generic response misses.
  • Accountability: keep a record of what was checked and take responsibility for the final result.

An AI tool might draft interview questions. A responsible recruiter still checks whether they are job-related, fair and free from inappropriate personal questions. It might suggest a maintenance checklist, but a qualified person must confirm the safe procedure.

Ask, "What would make this output trustworthy?" Then add that evidence.

Week 4: create proof without exaggerating

Build a one-page portfolio note containing:

  • The problem you were trying to solve.
  • The information you used.
  • How AI assisted.
  • What you checked manually.
  • What you changed and why.
  • The remaining limitations.

Use synthetic or public data and remove personal information. Do not claim invented productivity percentages. If you did not measure a time saving, say that the tool reduced first-draft effort or helped organise information.

A credible CV bullet could say: "Used an approved AI assistant to structure first-draft customer responses, then checked each response against policy and edited it for clarity." It describes a process without pretending the machine worked unsupervised.

In an interview, be prepared to explain a failure. Employers may learn more from how you caught an error than from a perfect demonstration.

Keep five safety rules

  • Follow the employer's AI and information-security policies.
  • Verify factual claims, calculations and citations.
  • Never present generated work as expert approval.
  • Protect personal and confidential information.
  • Keep a human accountable for decisions affecting people, money, safety or legal rights.

AI literacy is not a race to automate everything. It is the habit of using tools where they add value while strengthening the parts of work that require context, care and responsibility.

Start with one task this month. A small, well-checked improvement is better evidence than a long list of AI tools on your CV.

Sources and limits