Four AI Fluency Skills: Delegation, Description, Discernment, Diligence

Writing a good prompt is useful, but working well with AI also requires choosing the right task, checking the answer, and owning the result. The AI Fluency Framework, developed by Rick Dakan and Joseph Feller in collaboration with Anthropic, describes these habits as four connected competencies: Delegation, Description, Discernment, and Diligence.

What the four Ds mean in daily work

• Delegation: decide what an AI tool should do, what a person must do, and where review is needed.
• Description: give the tool a clear goal, relevant source material, constraints, and an output format.
• Discernment: evaluate the output against evidence and the actual task, then revise or reject it.
• Diligence: protect data, follow applicable policies, disclose AI involvement when required, and remain accountable for the work.

These competencies reinforce each other. A clear prompt cannot make an inappropriate task safe to delegate. A polished answer still needs checking. Good verification does not excuse sharing confidential material with an unapproved service.

1. Delegation: choose a bounded task

Start with a task that is reversible and easy to review: draft an outline, organize notes you are allowed to share, or suggest alternative headings. Keep final decisions with a person when a mistake could affect someone's money, safety, rights, or access. Define which source the tool may use and when a reviewer should stop the process. TipsMake has a practical guide to choosing suitable AI tasks.

2. Description: give a useful brief

For an approved task, describe the goal and audience, paste the relevant material, name hard constraints, and state what a good answer must include. Example: “Turn these approved meeting notes into a five-bullet update for the project team. Use only facts in the notes. Mark missing dates as ‘not stated.’ Do not send it.” The last sentence keeps drafting separate from publication.

3. Discernment: test the output

Check names, dates, numbers, and quotations against the original records. Ask whether the answer actually serves the reader and whether it omitted a caveat. If the model supplies a citation, open it and compare the claim with the source. A useful review log can mark each factual claim as supported, contradicted, or unresolved. See TipsMake's factual claim verification workflow.

4. Diligence: own the final action

Remove unnecessary personal or confidential data before prompting. Follow your organization's rules for approved tools and review. If a public deliverable needs disclosure or attribution, provide it. A person who is authorized to make the decision should sign off on consequential content or actions. Keep a record of what the tool drafted and what a person checked.

Build a small example of your own

1. Choose one recurring, low-risk task you already understand.
2. Write one sentence explaining why AI is suitable for a limited part of it.
3. Draft the prompt with source material, constraints, and output format.
4. List three checks you will perform on the result.
5. Record the final human decision: use, revise, or discard.

That one-page record demonstrates more than a claim that you “know AI.” It shows where AI helped, where it failed, and how you handled the result. Do not use a model's self-awarded score as proof of your own skill; judge the actual workflow.

Check your understanding

• Question 1:

  Why is it better to assess AI fluency through observed habits than self-confidence alone?

  • A. Confidence always proves that the output has been checked.
  • B. Confidence can exceed actual ability, while habits can be observed and improved.
  • C. The four competencies cannot be practiced.
  • D. The tool can automatically certify proficiency.

  EXPLAIN:

  A person can feel proficient without checking outputs or protecting data. Concrete examples of work reveal what was actually done and where the process needs improvement.

• Question 2:

  Which four competencies make up the 4D AI Fluency Framework?

  • A. Delegation, Description, Discernment, Diligence
  • B. Data, Design, Deployment, Debugging, and Documentation
  • C. Drafting, Dictating, Downloading, Deleting
  • D. Directing, Deciding, Detecting, Delivering/Completing

  EXPLAIN:

  The "4 D's" theoretical framework – Delegation, Description, Discernment, and Diligence – was developed by Professors Rick Dakan and Joseph Feller in collaboration with Anthropic. Each skill is a separate component that you can practice and improve; this course will focus on one skill in each lesson.

• Question 3:

  In this course's approach, what role does the skill of writing prompts play?

  • A. That's the whole concept of "AI proficiency" - simply mastering this skill is enough.
  • B. It is part of Description; the other competencies also matter.
  • C. It is one of four skills, equally important to the other three in the theoretical framework.
  • D. That is a distinct skill that this theoretical framework doesn't address at all.

  EXPLAIN:

  Prompt writing belongs to Description. The framework does not assign numerical weights to the four competencies; selecting tasks, evaluating results, and handling responsibility also matter.